{"id":"6c6857f7-e490-407f-afaa-761c84a97118","title":"Wikipedia Discussion Summary: Paid Translation and LLM Rewrites","ack":"2026-02-21T23:56:16Z","created_at":"2026-02-21T23:51:52Z","saved":true,"shared":true,"branch_id":"00000000-0000-4000-0000-000000000000","profile":{"id":"1b6d99f4-1649-48c8-8ee3-d6774318f814","name":null,"accessible":true,"model":"claude-4-6-opus-thinking","deprecate":false,"retired":false,"successor_model":null,"model_name":"Claude 4.6 Opus (reasoning)","model_provider":"anthropic","model_input_limit":1000000,"recommended":true,"model_info":"\n\n\n\n<div class=\"heading\">\n <div class=\"left\">\n \n \n \n <svg class=\"model-icon icon-xs\" data-model=\"claude-4-6-opus-thinking\">\n <use href=\"#icon-sparkles\"/>\n </svg>\n <h3>\n \n <span class=\"model-name\">none</span>\n </h3>\n </div>\n\n \n <a href=\"/settings?p=custom_assistant&id=1b6d99f4-1649-48c8-8ee3-d6774318f814\">\n <svg class=\"icon-sm edit-icon\" aria-hidden=\"true\">\n <use href=\"#icon-pencil\" />\n </svg>\n </a>\n \n</div>\n\n\n<hr />\n<ul>\n \n <li>\n <h4>\n Model\n </h4>\n <span class=\"flex align-center gap-8\">\n <svg class=\"icon-xs icon\">\n <use href=\"#icon-anthropic\" />\n </svg>\n <span class=\"model-name\">Claude 4.6 Opus (reasoning)</span>\n </span>\n </li>\n <li>\n <h4>\n Web access\n </h4>\n <span>On</span>\n </li>\n <li>\n <h4>\n Lenses\n </h4>\n <span></span>\n </li>\n \n</ul>\n\n\n","model_provider_name":"Anthropic","internet_access":true,"personalizations":false,"lens":{"id":21245,"name":null},"shortcut":null,"is_default_profile":false},"tag_ids":[],"read_only":true}

[{"id":"1d5b73d7-0ede-480c-80de-32c8bd5c9e8d","thread_id":"6c6857f7-e490-407f-afaa-761c84a97118","created_at":"2026-02-21T23:51:52Z","branch_list":["00000000-0000-4000-0000-000000000000","44f1175b-fd20-412b-9638-63443c026b2a"],"state":"done","prompt":"Summarise the arguments in the discussion thread at <https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.>, and determine whether there is any clear consensus or call to action.","references_html":"<ol data-ref-list>\n \n <li>\n <a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators&#39;_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.\" >Wikipedia:Administrators&#39; noticeboard - Wikipedia</a>\n <span class=\"__domain-name\">en.wikipedia.org</span>\n <div class=\"trigger\" tabindex=\"0\">\n <span class=\"contribution\" style=\"background: linear-gradient(to right, var(--primary-70) 100%, transparent 100%)\">100%</span>\n <div class=\"tooltip top end tooltip_base tw-auto\">\n This percentage indicates how much a source has contributed to the answer.\n </div>\n </div>\n </li>\n \n</ol>\n","references_md":"[^1]: [Wikipedia:Administrators' noticeboard - Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.) (100%)\n","reply":"<details><summary>Gathered key details from:<p></p>\n<ul>\n<li><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.\" target=\"_blank\">Wikipedia:Administrators' noticeboard - Wikipedia</a></li></ul></summary>\n\n<h3>Query</h3>\n<p>Summarise the arguments in the discussion thread and determine if there is any clear consensus or call to action.</p>\n<h3>Response</h3>\n<p>The discussion is a long ANB/AN thread about OKA translation project. Key themes: concerns about AI/LLM use, quality control, training, transparency, funding, onboarding, supervision, and potential conflicts of interest. Main points raised: 1) LLM hallucinations and accuracy; 2) Overwrites and content quality; 3) Mechanisms to improve QA: vetting, peer review, warnings, disruptions; 4) Onboarding, compensation, independence vs. paid editing; 5) Targeted donations and COI disclosures; 6) Calls for more transparency about governance and donors; 7) Calls for policy changes (RFCs, AFC, AFCSTANDARDS) and potential sanctions; 8) Specific incidents with OKA; 9) Broader governance questions about consensus on content decisions (BLP, NOTCENSORED, V, ONUS). There is no single clear consensus; opinions vary from strong opposition to cautious support and conditional proposals. Some recurring calls to action: implement a four-strikes policy for translators with warnings and potential blocking; pilot peer-review and increased oversight; publish more on-wiki governance and donor disclosures; potentially start RFCs or BLPN discussions; consider restricting OKA translators from direct mainspace edits; and continue data gathering and transparent reporting. The thread ends with ongoing debates and no definitive consensus; some participants urge continuing discussion through RFCs and BLPN, etc. Overall: no firm consensus; some calls to action include implementing stricter oversight, transparency, and possibly restricting paid editing while others advocate measured changes and more data. End.</p>\n</details><details><summary>Thinking</summary>\n<p>The user is asking me to summarise a discussion thread on the English Wikipedia Administrators' noticeboard about an \"OKA\" translation project and its use of LLMs. Let me work with the information provided to give a structured summary.</p>\n</details>\n<h2>Summary: AN thread on OKA paid translation and LLM-assisted editing</h2>\n<p>The thread concerns the <strong>OKA translation project</strong>, which appears to involve paid translators making large-scale lead rewrites and translations across thousands of articles, with concerns that LLMs (large language models) are being used in the process. Below are the main lines of argument.</p>\n<h3>Key arguments raised</h3>\n<table>\n<thead>\n<tr>\n<th>Theme</th>\n<th>Summary</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td><strong>LLM hallucinations &amp; accuracy</strong></td>\n<td>Editors flag that LLM-generated or LLM-assisted content introduces factual errors, unsourced claims, and subtle hallucinations that are difficult to catch at scale.</td>\n</tr>\n<tr>\n<td><strong>Overwrites &amp; content quality</strong></td>\n<td>Concern that existing, human-written lead sections are being overwritten with lower-quality or less policy-compliant text, sometimes degrading BLP-sensitive articles.</td>\n</tr>\n<tr>\n<td><strong>Paid editing &amp; COI transparency</strong></td>\n<td>Questions about whether OKA translators are adequately disclosing their paid status per the Terms of Use and English Wikipedia's paid-editing disclosure requirements, and whether targeted donations create a conflict of interest.</td>\n</tr>\n<tr>\n<td><strong>Onboarding &amp; supervision</strong></td>\n<td>Debate over whether translators are sufficiently trained in core content policies (WP:V, WP:BLP, WP:ONUS, WP:NPOV) before being given access to edit mainspace at volume.</td>\n</tr>\n<tr>\n<td><strong>Governance &amp; donor disclosure</strong></td>\n<td>Calls for more on-wiki transparency about who funds the project, how it is governed, and how quality assurance decisions are made.</td>\n</tr>\n<tr>\n<td><strong>Policy mechanisms</strong></td>\n<td>Various proposals floated: stricter AFC-style gatekeeping, a \"four-strikes\" warning-and-block escalation for translators, restricting OKA accounts from direct mainspace edits, and launching formal RFCs or BLPN discussions.</td>\n</tr>\n<tr>\n<td><strong>Measured / supportive voices</strong></td>\n<td>Some participants argue the project has genuine value (expanding coverage in under-represented languages), advocate for incremental improvements and more data rather than blanket restrictions, and caution against over-reaction.</td>\n</tr>\n</tbody>\n</table>\n<h3>Consensus and calls to action</h3>\n<p>There is <strong>no single clear consensus</strong> in the thread <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=The%20discussion%20is,more%20data.%20End.\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=The%20discussion%20is,more%20data.%20End.\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">The discussion is a long ANB/AN thread about OKA translation project. Key themes: concerns about AI/LLM use, quality control, training, transparency, funding, onboarding, supervision, and potential conflicts of interest. Main points raised: 1) LLM hallucinations and accuracy; 2) Overwrites and content quality; 3) Mechanisms to improve QA: vetting, peer review, warnings, disruptions; 4) Onboarding, compensation, independence vs. paid editing; 5) Targeted donations and COI disclosures; 6) ...</span></span></sup>. Opinions range from strong opposition (calling for immediate restrictions or blocks) to cautious support contingent on better oversight. However, several <strong>recurring calls to action</strong> can be identified:</p>\n<ul>\n<li><strong>Stricter quality oversight</strong> — implement structured peer review and a graduated warning/blocking policy for translators who introduce errors.</li>\n<li><strong>Greater transparency</strong> — publish on-wiki governance details, donor information, and quality-assurance metrics.</li>\n<li><strong>Restrict direct mainspace editing</strong> — some editors want OKA translators to route edits through a review process (analogous to AFC) rather than editing mainspace directly.</li>\n<li><strong>Formal community processes</strong> — multiple participants urge moving the discussion into RFCs, BLPN threads, or dedicated policy proposals so that binding community consensus can be established.</li>\n<li><strong>Continued data gathering</strong> — calls to quantify the scale of errors before deciding on sanctions.</li>\n</ul>\n<p>In short, the thread surfaces serious and widely shared concerns about quality, disclosure, and governance, but the community has not yet coalesced around a specific remedy; the most common next step suggested is to formalise the discussion through RFCs or other structured processes <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=The%20discussion%20is,more%20data.%20End.\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=The%20discussion%20is,more%20data.%20End.\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">The discussion is a long ANB/AN thread about OKA translation project. Key themes: concerns about AI/LLM use, quality control, training, transparency, funding, onboarding, supervision, and potential conflicts of interest. Main points raised: 1) LLM hallucinations and accuracy; 2) Overwrites and content quality; 3) Mechanisms to improve QA: vetting, peer review, warnings, disruptions; 4) Onboarding, compensation, independence vs. paid editing; 5) Targeted donations and COI disclosures; 6) ...</span></span></sup>.</p>","md":"<details><summary>Thinking</summary>\n\nThe user is asking me to summarise a discussion thread on the English Wikipedia Administrators' noticeboard about an \"OKA\" translation project and its use of LLMs. Let me work with the information provided to give a structured summary.\n\n</details>\n\n## Summary: AN thread on OKA paid translation and LLM-assisted editing\n\nThe thread concerns the **OKA translation project**, which appears to involve paid translators making large-scale lead rewrites and translations across thousands of articles, with concerns that LLMs (large language models) are being used in the process. Below are the main lines of argument.\n\n### Key arguments raised\n\n| Theme | Summary |\n|---|---|\n| **LLM hallucinations & accuracy** | Editors flag that LLM-generated or LLM-assisted content introduces factual errors, unsourced claims, and subtle hallucinations that are difficult to catch at scale. |\n| **Overwrites & content quality** | Concern that existing, human-written lead sections are being overwritten with lower-quality or less policy-compliant text, sometimes degrading BLP-sensitive articles. |\n| **Paid editing & COI transparency** | Questions about whether OKA translators are adequately disclosing their paid status per the Terms of Use and English Wikipedia's paid-editing disclosure requirements, and whether targeted donations create a conflict of interest. |\n| **Onboarding & supervision** | Debate over whether translators are sufficiently trained in core content policies (WP:V, WP:BLP, WP:ONUS, WP:NPOV) before being given access to edit mainspace at volume. |\n| **Governance & donor disclosure** | Calls for more on-wiki transparency about who funds the project, how it is governed, and how quality assurance decisions are made. |\n| **Policy mechanisms** | Various proposals floated: stricter AFC-style gatekeeping, a \"four-strikes\" warning-and-block escalation for translators, restricting OKA accounts from direct mainspace edits, and launching formal RFCs or BLPN discussions. |\n| **Measured / supportive voices** | Some participants argue the project has genuine value (expanding coverage in under-represented languages), advocate for incremental improvements and more data rather than blanket restrictions, and caution against over-reaction. |\n\n### Consensus and calls to action\n\nThere is **no single clear consensus** in the thread [^1]. Opinions range from strong opposition (calling for immediate restrictions or blocks) to cautious support contingent on better oversight. However, several **recurring calls to action** can be identified:\n\n- **Stricter quality oversight** — implement structured peer review and a graduated warning/blocking policy for translators who introduce errors.\n- **Greater transparency** — publish on-wiki governance details, donor information, and quality-assurance metrics.\n- **Restrict direct mainspace editing** — some editors want OKA translators to route edits through a review process (analogous to AFC) rather than editing mainspace directly.\n- **Formal community processes** — multiple participants urge moving the discussion into RFCs, BLPN threads, or dedicated policy proposals so that binding community consensus can be established.\n- **Continued data gathering** — calls to quantify the scale of errors before deciding on sanctions.\n\nIn short, the thread surfaces serious and widely shared concerns about quality, disclosure, and governance, but the community has not yet coalesced around a specific remedy; the most common next step suggested is to formalise the discussion through RFCs or other structured processes [^1].","profile":{"id":"1b6d99f4-1649-48c8-8ee3-d6774318f814","name":"wikimedia-policy","accessible":true,"model":"claude-4-6-opus-thinking","deprecate":false,"retired":false,"successor_model":null,"model_name":"Claude 4.6 Opus (reasoning)","model_provider":"anthropic","model_input_limit":1000000,"recommended":true,"model_info":"\n\n\n\n<div class=\"heading\">\n <div class=\"left\">\n \n \n \n <svg class=\"model-icon icon-xs\" data-model=\"claude-4-6-opus-thinking\">\n <use href=\"#icon-sparkles\"/>\n </svg>\n <h3>\n \n <span class=\"model-name\">wikimedia-policy</span>\n </h3>\n </div>\n\n \n <a href=\"/settings?p=custom_assistant&id=1b6d99f4-1649-48c8-8ee3-d6774318f814\">\n <svg class=\"icon-sm edit-icon\" aria-hidden=\"true\">\n <use href=\"#icon-pencil\" />\n </svg>\n </a>\n \n</div>\n\n\n<hr />\n<ul>\n \n <li>\n <h4>\n Model\n </h4>\n <span class=\"flex align-center gap-8\">\n <svg class=\"icon-xs icon\">\n <use href=\"#icon-anthropic\" />\n </svg>\n <span class=\"model-name\">Claude 4.6 Opus (reasoning)</span>\n </span>\n </li>\n <li>\n <h4>\n Web access\n </h4>\n <span>On</span>\n </li>\n <li>\n <h4>\n Lenses\n </h4>\n <span></span>\n </li>\n \n</ul>\n\n\n","model_provider_name":"Anthropic","internet_access":true,"personalizations":false,"lens":{"id":21245,"name":null},"shortcut":null,"is_default_profile":false},"metadata":"<li>\n <span class=\"attribute\">Model</span>\n <span class=\"value\"><i class=\"icon-xs mr-4\"><svg><use href=\"#icon-anthropic\"></use></svg></i>Claude 4.6 Opus (reasoning)</span>\n</li>\n<li>\n <span class=\"attribute\">Version</span>\n <span class=\"value\">claude-opus-4-6</span>\n</li>\n<li>\n <span class=\"attribute\">Speed (tok/s)</span>\n <span class=\"value\">39</span>\n</li>\n<li>\n <span class=\"attribute\">Tokens</span>\n <span class=\"value\">3,124</span>\n</li>\n<li>\n <span class=\"attribute\">Cost / Total ($)</span>\n <span class=\"value\">0.05 / 0.05</span>\n</li>\n<li>\n <span class=\"attribute\">End to end time (s)</span>\n <span class=\"value\">67.38</span>\n</li>\n<li>\n <span class=\"attribute\">Submitted</span>\n <span class=\"value\">2026-02-21 18:51</span>\n</li>\n\n","documents":[]},{"id":"baae8118-e56b-4ece-bc06-a7cbf8aa24d5","thread_id":"6c6857f7-e490-407f-afaa-761c84a97118","created_at":"2026-02-21T23:56:15Z","branch_list":["00000000-0000-4000-0000-000000000000"],"state":"done","prompt":"Identify the editors who have contributed to this discussion, and summarise their positions.","references_html":"<ol data-ref-list>\n \n <li>\n <a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators&#39;_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.\" >Wikipedia:Administrators&#39; noticeboard - Wikipedia</a>\n <span class=\"__domain-name\">en.wikipedia.org</span>\n <div class=\"trigger\" tabindex=\"0\">\n <span class=\"contribution\" style=\"background: linear-gradient(to right, var(--primary-70) 100%, transparent 100%)\">100%</span>\n <div class=\"tooltip top end tooltip_base tw-auto\">\n This percentage indicates how much a source has contributed to the answer.\n </div>\n </div>\n </li>\n \n</ol>\n","references_md":"[^1]: [Wikipedia:Administrators' noticeboard - Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.) (100%)\n","reply":"<details><summary>Gathered key details from:<p></p>\n<ul>\n<li><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.\" target=\"_blank\">Wikipedia:Administrators' noticeboard - Wikipedia</a></li></ul></summary>\n\n<h3>Query</h3>\n<p>Summarise the arguments and identify the editors who have contributed to the discussion thread.</p>\n<h3>Response</h3>\n<h2>OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles.</h2>\n<p>\n</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>e</mi><mi>d</mi><mi>i</mi><mi>t</mi></mrow></math></div>\n<p></p>\n<h3>Context</h3>\n<p>\n</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>e</mi><mi>d</mi><mi>i</mi><mi>t</mi></mrow></math></div>\n<p></p>\n<ul>\n<li>OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration.</li>\n<li>OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. They're explicitly encouraging use of general-purpose LLMs for translation.</li>\n<li>Their public Google Sheet tracker includes a tab <em>Simplify lead</em> with instructions directing editors to Instructions: pick an article, copy the lead section into Gemini or chatGPT, then review if some of the suggestions are an improvement to readability. Make edits to the Wiki articles only if the suggestions are an improvement and don't change the meaning of the lead. Do not change the content unless you have checked that what Gemini says is correct!</li>\n</ul>\n<h3>Existing discussions/reports</h3>\n<p>\n</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>e</mi><mi>d</mi><mi>i</mi><mi>t</mi></mrow></math></div>\n<p></p>\n<ul>\n<li>October 2022 at WP:VPM: OKA Founder User:7804j announces the formation of the non-profit. Early responses showed machine-translation errors, poor English competency, duplicate articles, overproducing low-quality pages that volunteers then had to clean up. <ul>\n<li>many of the results are very problematic, with a large number of WP:CFORKs, editors who clearly have very poor English, don't read through their work (or are incapable of seeing problems) and don't add links and so on. As the OKA instructions linked above make clear, these are machine translations, with all the problems that brings. link </li>\n<li>the translations include duplicates of topics we already have link </li>\n<li>My understanding is that professional translation usually requires natives of the target language for the best result and so the OKA might be going against the natural flow. Apart from command of the language, there are also social factors to consider. The English language Wikipedia is rife with conflict and even native English speakers are often discouraged or driven off. link</li>\n</ul>\n</li>\n<li>November 2022 at WP:VPI: Suggested tighten controls on translations with flagging of low-quality translations by OKA editors with inadequate quality control and translator English proficiency. <ul>\n<li>In recent months I'm seeing a number of translations, from Spanish and Portuguese, of long and important articles on art history that might or not be machine translations (probably they are) but are certainly not checked by anyone competent in English link</li>\n</ul>\n</li>\n<li>November 2023 at WP:COIN: OKA editors producing new articles containing lavish tangential detail that might function as subtle promotion, with machine-translation artefacts. <ul>\n<li>The church article is a way too faithful translation of the German one, without a look at that English church articles should look different ... I would not know where to start to bring the product in line with the English Wikipedia ... articles like that add work to reviewers, perhaps more than they provide facts to readers. link </li>\n<li>The trouble is, they are machine translations by people often with a really poor level of English. They are not adding new content, just translating the whole of the German or whatever article they've chosen. If they bother to do any checking through, which mostly they don't seem to, their English is too poor to spot even glaring and basic mistakes. The linking is always terrible. link</li>\n</ul>\n</li>\n<li>May 2024 at OKA Founder User:7804j: OKA editors not following COI or PAID disclosures. <ul>\n<li>I have looked through some of the editors from this, and they are not following the conflict of interest or paid editing requirements.</li>\n</ul>\n</li>\n<li>May 2024 at WP:Intertranswiki: OKA Founder User:7804j queried why OKA editors had to go via AfC to create new articles. <ul>\n<li>OKA creates several thousand new articles per year (99% of which are translations), so I have concerns that if all of these were to go through AFC, it would completely clog the process and makes us unable to operate. We will test it but, given the circumstances, there is a real risk we may need to shut down OKA entirely on EN WP and re-assign our \\~15 full-time translators to other Wikipedia languages with less stringent paid editing policies link</li>\n</ul>\n</li>\n<li>June 2024 at WP:VPP: A discussion on if OKA editors should go through AfC. Cited recurring problems like minimally edited machine-translation prose, awkward titles, poor linking/categorization, no aftercare, and content forks/duplication with AfC justified to protect enwiki from extra cleanup burden. <ul>\n<li>Our core principle is that our translators are free to work on anything that interests them. We provide them with a monthly stipend, some training on how Wikipedia works, but we then see them as volunteer contributors on whom we impose some process to ensure they do not abuse the grant and provide overall value (eg, quality checks, quantity checks). To help them find articles to translate, we curate an optional backlog ... Articles of this tracker primarily consist of \"Featured\" and \"Good\" quality articles from other Wikis, as well as red links from these articles </li>\n<li>We're not paying per quantity, but per hour of work and instructing that people should focus on quality. Our translators are also paid when they work on improvements of existing articles. </li>\n<li>They have caused a lot of work; mostly these are machine translations by people whose English is rather poor. </li>\n<li>I've seen no evidence that OKA translators are creating unedited machine translations. </li>\n<li>7804j has never denied that these are machine translations, and they normally appear on en:wp in a single edit, &amp; are not edited further except for a couple of tidies. There is no evidence that they are edited machine translations when OKA bow out, and they should be treated as \"unedited machine translations\"</li>\n</ul>\n</li>\n<li>September 24 at Talk:Lycée Molière (Paris): An article translated into English by an OKA editor using LLM, with quality issues. <ul>\n<li>The school's significance gets lost among the minutiae of historical enrolment numbers, timetables, uniforms, and report cards – details that might just be acceptable in an institutional history published by and for the school, but have no place in an encyclopaedia which is supposed to summarise sources, not reproduce trivia from the school's newsletter archive.</li>\n</ul>\n</li>\n<li>01 January 2026 at WP:ANI: An OKA editor repeatedly broke infoboxes, large translation overwrites that introduced MOS issues, and failure to communicate. Temporarily blocked, then unblocked and resolved after editor promised to be more careful. <ul>\n<li>I have also twice reverted their edits with a summary explaining that communication is required and that they are breaking the page </li>\n<li>article edits all appear to be expansions by overlaying all or most of the article with a translation</li>\n</ul>\n</li>\n<li>02 January 2026 at WP:AINB: Unreviewed / poorly reviewed AI translations by an OKA editor. Unresolved. <ul>\n<li>it is clear that the rich in preserve a rich and varied folklore from Acadian folklore isn't in the original frwiki text (and neither is the *relevailles* markdown). Not looking good</li>\n</ul>\n</li>\n<li>07 January 2026 at WP:AINB: Unreviewed / poorly reviewed AI translations with synthesis and promo-language by an OKA editor. Unresolved. <ul>\n<li>There is AI text that does not have any counterpart in it, most of which are synthesis tacked on to paragraphs</li>\n</ul>\n</li>\n</ul>\n<h4>The ongoing VPP thread which prompted me to create this AN report</h4>\n<p>\n</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>e</mi><mi>d</mi><mi>i</mi><mi>t</mi></mrow></math></div>\n<p></p>\n<ul>\n<li>January 2026 at WP:VPP: The thread starts with OKA Founder User:7804j asking whether the new WP:NEWLLM policy also effectively bans LLM-assisted translations that are, which would disrupt the OKA model. Also includes an example of a OKA translated article containing info not verified by the source. <ul>\n<li>AI translations do not tend to be faithful to the original, and the text that deviates tends to contain the usual problems with standard AI-generated text. In practice the level of review being done is far short of that. For example, this \"translation\" inserted an entire chunk of text under \"River transport\" when the corresponding section in the French article was completely empty. </li>\n<li>The level of human review to AI additions is, quite often, very insufficient </li>\n<li>However, looking at the provided page number, 147, it doesn't talk about the La Bourdonnaye family at all – instead, the families listed on that page range from Cadelac to Cado.The actual mention of Bourdonnaye, on page 114, does not talk at all about them originating from a lordship in Trégomar, Côtes-d'Armor. I am worried that you might not have actually checked the sources that you have been adding, and I would be happy to see evidence of the contrary.</li>\n</ul>\n</li>\n</ul>\n<h3>Main concerns</h3>\n<p>\n</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>e</mi><mi>d</mi><mi>i</mi><mi>t</mi></mrow></math></div>\n<p></p>\n<ul>\n<li><em>Edit quality</em>: OKA editors are making large-scale changes that introduce errors and LLM artefacts.</li>\n<li><em>Overwrites</em>: repeated overwrites of existing articles during LLM-assisted translation which breaks templates/infoboxes, and makes it hard to review because they're not incremental changes.</li>\n<li><em>Coordinated mass-editing</em>: OKA trackers indicate various planned campaigns for mass changes (e.g. the Simplify lead campaign) affecting large numbers of pages without on-wiki discussion or consensus.</li>\n<li><em>QA</em>: from the outside, it’s not clear what OKA’s actual QA looks like (vetting, onboarding, supervision), given the above outcomes.</li>\n</ul>\n<h3>Suggested remedies</h3>\n<p>\n</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>e</mi><mi>d</mi><mi>i</mi><mi>t</mi></mrow></math></div>\n<p></p>\n<ul>\n<li><em>Edit quality &amp; Overwrites:</em> OKA editors using LLMs to translate must check for quality and accuracy (also pending an LLM translation policy RFC). OKA editors performing translation expansions on existing articles must use incremental edits rather than text overwrites.</li>\n<li><em>Coordinated mass-editing:</em> Editing campaigns should either have 1) on-Wiki consensus on the affected articles before edits go ahead or 2) required to submit their proposed changes via Edit Requests on the article Talk page.</li>\n<li><em>QA:</em> OKA to provide documentation regarding their hiring, onboarding, and supervision processes. How do they assess English competency? How they check that LLM output hasn’t added unsupported content? etc</li>\n</ul>\n<p>Courtesy ping: @7804j as OKA Founder. <strong>qcne</strong> (talk) 14:54, 19 January 2026 (UTC)\\[\\] </p>\n<p>This is a long-running and serious problem (some of the ye olde quotes above are by me) &amp; while the suggested remedies would be fine if they were made to work, it has long been clear that 7804j is only intermittently interested in engaging with the community, and puts very little effort at all into quality control over the people he pays. He is not himself an en:wp editor with any real experience, and simply does not understand many of the issues. One key issue is identifying OKA-paid material. There used to be a speadsheet that did this, but I don't know if this is still accessible. Most, but I think not all, the editors he pays identify themselves on their user or talk pages, but the articles they create are very often not given any sort of tag. Many of them have pretty poor English, so must use machine-translation of articles from other WPs, or I suppose now freshly-brewed AI texts. A remedy like \"OKA editors using LLMs to translate must check for quality and accuracy\" just won't fly with the existing set-up - the editors just won't know how to do this. I haven't noticed that the situation common in the early days persists, where an OKA editor set up a FORK for a major article evidently without checking whether en:wp already had one (normally we did, with a rather more idiomatic title). He was warned about that, &amp; it may have stopped - the solution was just to redirect the OKA title to the existing one, so his money was wasted. Johnbod (talk) 15:26, 19 January 2026 (UTC)\\[\\]</p>\n<p>To me this comment \"</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mo>.</mo><mo>.</mo><mo>.</mo></mrow></math></div> we may need to shut down OKA entirely on EN WP and re-assign our \\~15 full-time translators to other Wikipedia languages <div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mo>.</mo><mo>.</mo><mo>.</mo></mrow></math></div>\" shows that they are indeed using machine translation. I don't see how a translator could translate competently into English, and then just switch at the drop of a hat to translating into Finnish or Farsi instead.<p></p>\n<p>I have accepted a few of these at AfC, and they seemed quite okay to me, although I admit I didn't specifically look for possible MT problems. I also feel, rightly or wrongly, like investigating the translation quality or provenance would be beyond the scope of AfC's (or for that matter, NPP's) remit. -- DoubleGrazing (talk) 15:55, 19 January 2026 (UTC)\\[\\] </p>\n<p>If he meant any random language, sure. But a lot of translators can do bi-directional translation, so if you're doing es→en, it might be easy to take assignments that go en→es. Many translators know multiple languages, too. (For that matter, my plumber speaks seven languages. In some parts of the world, being multilingual isn't remarkable.) WhatamIdoing (talk) 23:56, 22 January 2026 (UTC)\\[\\] </p>\n<p>Fluency is also a spectrum. I know enough French to watch movies and read books, which is useful when there's an annoying lack of English sources for something. But my French conversational skills are severely lacking and my writing is iffy on a grammatical level. There's a reason I don't hang out on fr.wiki even if I read their articles sometimes. Clovermoss🍀 (talk) 21:11, 23 January 2026 (UTC)\\[\\] </p>\n<p>Yes; similar for me. And sometimes the limited fluency leads to one-way translation: I'm confident translating from French to English, but less so from English to French. Cremastra (talk <strong>·</strong> contribs) 23:09, 23 January 2026 (UTC)\\[\\]</p>\n<p>A significant concern that occurred during the NEWLLM discussion was this edit, where an obviously llm-generated text with fabricated sources was ~~restored~~detagged by 7804j. While it was later removed, I do note checking that a fabricated source added elsewhere in the article was not removed. I have removed it now, but clearly the issue extended beyond just that additional portion and I doubt anyone has actually checked the other sources to see if they were fabricated. I don't know if we need an OKA specific remedy, but hopefully OKA participants are made aware of the potential issues they might run into on en.wiki. CMD (talk) 17:13, 19 January 2026 (UTC)\\[\\]</p>\n<p>I appreciate the community bringing these concerns forward. I recognize that as OKA grows, our activity becomes more visible. I want to address the specific points raised and clarify how we manage quality.</p>\n<p>Regarding OKA’s output, it is essential to look at the data in context. With over 13,000 articles created globally (including 4,800 on English Wikipedia) by more than 80 OKA editors, a small number of errors are statistically inevitable in any high-volume project, whether volunteer or funded. While we treat every report of a problematic edit as an opportunity to refine our internal guidance, these instances remain outliers. They do not represent the vast majority of our work, which provides a significant net benefit by filling content gaps and expanding articles. Our goal is not to \"clog the process\" but to continue optimizing our internal checks so that our positive impact remains high while minimizing the burden on other editors.</p>\n<p><strong>LLM usage and accuracy</strong></p>\n<p>The concerns regarding fabricated sources (hallucinations) are the most serious.</p>\n<ul>\n<li><strong>Source verification:</strong> We are reinforcing to all grant recipients that LLMs are tools for phrasing and translation, not for facts. Any editor found adding unsupported content or fabricated citations receives a formal warning.</li>\n<li><strong>Policy alignment:</strong> I initiated the recent discussion that led to the current RFC because I want OKA to work within the community’s consensus on WP:NEWLLM. We are also finalizing a study on AI effectiveness to share with the community.</li>\n</ul>\n<p><strong>Process and vetting</strong></p>\n<ul>\n<li><strong>Overwrites:</strong> I fully agree with the feedback here. Overwriting existing content is not ok, unless that content is poor. This is clearly reflected in our instructions. Any editor who ignores this receives a formal warning; two warnings result in the immediate suspension of their OKA grant.</li>\n<li><strong>English proficiency:</strong> Our process already requires editors to demonstrate their ability to produce idiomatic, high-quality English that meets WP:MOS before they are eligible for a stipend. We aren't \"hiring\" staff, but rather providing grants to independent contributors who meet these standards.</li>\n<li><strong>Transparency:</strong> Our tracker is public at oka.wiki/tracker (and on-wiki). Every article we work on is tagged as such in the talk page too. We are open to suggestions on how to make OKA-funded edits more easily identifiable (e.g., specific edit summaries) to assist reviewers.</li>\n</ul>\n<p><strong>Clarifications on specific comments</strong></p>\n<ul>\n<li><strong>On my experience:</strong> Johnbod mentioned I am not myself \"an en:wp editor with any real experience\". This is not true -- I have been a Wikipedia editor for over 10 years. Like many long-term contributors, my focus has been primarily on the content side rather than policy discussions or noticeboards, so it is true I have less experience on these aspects. I am engaging here to ensure the community has the facts about our workflow.</li>\n<li><strong>On \"switching languages\":</strong> To clarify: our editors only work in languages they are fluent in. My previous comment about \"re-assigning to other languages\" referred to moving our funding focus to different language editions of Wikipedia, not asking individual translators to work in languages they don't know.</li>\n<li><strong>The \"fabricated source\" incident:</strong> To be clear, I did not \"restore\" fabricated text. I removed a maintenance tag based on a misunderstanding of whether the section was a translation or an expansion. Once it became clear the editor hadn't verified the sources properly, we took action. I'll take responsibility for the mistake in that specific review, but it was an isolated error.</li>\n</ul>\n<p><strong>Moving forward</strong></p>\n<p>We don't want OKA editors to be held to a \"different\" standard, but we recognize that as a funded organization, we have a responsibility to ensure we aren't creating extra work for the community. We believe the current model provides a substantial net gain for Wikipedia. 7804j (talk) 19:41, 19 January 2026 (UTC)\\[\\] </p>\n<p>Cmon man, the training guide instructs translators to create multiple email accounts to get around LLM usage caps… ExtantRotations (talk) 20:12, 19 January 2026 (UTC)\\[\\] </p>\n<p>...yes, and? 7804j (talk) 20:16, 19 January 2026 (UTC)\\[\\]</p>\n<p>Did you write this with the usage of an LLM, @7804j? Three AI detectors suggest it's AI generated. <strong>qcne</strong> (talk) 21:08, 19 January 2026 (UTC)\\[\\] </p>\n<p>Whether or not I used an LLM, spell-checking tool, or the help of someone to review and provide feedback on my response is irrelevant. I would like us to focus on the substance of this discussion.</p>\n<p>Also keep in mind that tools that \"detect LLMs\" look for signs such as the presence of bullet points, headers, bolding, etc. which are also best practices for communication. Structuring arguments as lists and highlighting core ideas is always how I communicate when needing to convey complex ideas to a large audience, when I can spend sufficient time to copyedit and review my response. For example, the message I am currently writing is less structured (no bullets or headers), mostly because I am writing while I am on the go, thus having less time to iterate on it. 7804j (talk) 08:55, 20 January 2026 (UTC)\\[\\] </p>\n<p>The tools used while editing Wikipedia are directly relevant to this discussion. They are the very topic of this discussion. Being deliberately evasive on very simple questions is not a helpful way to develop trust. CMD (talk) 10:06, 20 January 2026 (UTC)\\[\\]</p>\n<p>@7804j It is absolutely relevant, and this response is disappointing. <strong>qcne</strong> (talk) 11:00, 20 January 2026 (UTC)\\[\\]</p>\n<p>We have ended up here at AN largely because of some OKA editors' careless use of LLMs, so yes, whether the founder and main driver of the project uses LLMs to communicate with us is very relevant. And having spent months working alongside Qcne at AFC, I am confident that he understands the nuances amd pitfalls of AI detection and isn't making this accusation lightly. <strong>ClaudineChionh</strong> (<em>she/her</em> · talk · email · global) 09:14, 20 January 2026 (UTC)\\[\\] </p>\n<p>Yes, I use LLMs to help me draft and structure my thoughts, especially for long responses where I want to be clear and address multiple points. I don’t believe there is anything inherently wrong with using modern tools to improve the quality and readability of a discussion, provided the person posting takes full responsibility for the content.</p>\n<p>I’m sorry if my previous response seemed evasive; that wasn’t my intent. My point was that the substance of the arguments should matter more than the software used to assist in writing them. To me, using an LLM to refine a post is a matter of personal preference and productivity, similar to using a sophisticated spell-checker or a translation tool.</p>\n<p>Trust is a two-way street. Being questioned on the mechanics of how I write feels like a distraction from the actual policy issues we are here to solve. Many professional environments now encourage the use of these tools specifically because they can improve the clarity of communication.</p>\n<p>The core issue here is whether OKA's output on Wikipedia is accurate and valuable. I am here to take responsibility for that output and to work on the process improvements I mentioned above. I would prefer we focus on those specific quality controls rather than the tools I use to draft my replies. 7804j (talk) 11:14, 20 January 2026 (UTC)\\[\\] </p>\n<p>I mean technically you aren’t taking responsibility for the output though; you have made it clear that all errors are the fault of your (quite low paid) employees, and that they will receive all the punishment. Nowhere in your list of changes do you discuss introducing additional training or help for translators. ExtantRotations (talk) 12:20, 20 January 2026 (UTC)\\[\\] </p>\n<p>I want to be clear about what I mean by responsibility. As the founder, I am responsible for the system and instructions our editors follow. If there is a pattern of errors, it means our training or oversight needs adjustment, and that is on me.</p>\n<p>When I say editors are responsible for their output, I’m simply stating the Wikipedia rule: every person is accountable for the edits they make. However, we don't just \"wait for mistakes to punish.\" Our responsibility is in the support we provide:</p>\n<ul>\n<li><strong>Training:</strong> We don't just hand out money; we provide onboarding and direct mentorship from experienced editors.</li>\n<li><strong>Systemic Fixes:</strong> When the community flags an issue—like overwriting or infobox errors—we don't just blame an individual. We update our global guidelines and hold group sessions to ensure every OKA editor understands the fix.</li>\n<li><strong>Quality Control:</strong> The \"warning\" system is a safeguard for Wikipedia. It’s there to ensure that if someone consistently fails to meet community standards despite our training, they stop receiving funding.</li>\n</ul>\n<p>I am not \"passing the buck.\" I am here to ensure that OKA as a whole is a net positive, which means I am constantly iterating on our training to prevent these issues from happening in the first place. 7804j (talk) 12:43, 20 January 2026 (UTC)\\[\\] </p>\n<p>&gt; We don't just hand out money; we provide onboarding</p>\n<p>Are the editors compensated/paid for their time during onboarding?</p>\n<p>As a personal request; if you reply to me, can you please use <em>only</em> your own words? I do not like talking to someone through a computer-generated layer, it actually feels disrespectful to me. --Gurkubondinn (talk) 12:57, 20 January 2026 (UTC)\\[\\] </p>\n<p>Yes, they are compensated during onboarding regardless of the outcome of the trial period 7804j (talk) 16:06, 20 January 2026 (UTC)\\[\\] </p>\n<p>Thanks a lot! I presume that this refers to a broader program than the initial 40 hour trial, as, regarding the latter, the job posting template states that:</p>\n<p>&gt; Note that payment for the trial requires that you publish at least 5 articles successfully without quality issues and share your progress daily within the established trial period.</p>\n<p>If you do refer to a longer paid onboarding, I would be happy to hear details about how it is being managed by the senior editors, and what oversight is present regarding this process. This could really help convince the editing community that your translators receive appropriate supervision and training. Chaotic Enby (talk · contribs) 16:18, 20 January 2026 (UTC)\\[\\] </p>\n<p>This sounds like a contraction to compensated during onboarding regardless of the outcome of the trial period. --Gurkubondinn (talk) 16:25, 20 January 2026 (UTC)\\[\\] </p>\n<p>The difference is that the job posting refers to the payment as requiring articles without quality issues, which isn't the same as regardless of the outcome, which is why I believe they must be referring to different processes. Chaotic Enby (talk · contribs) 16:38, 20 January 2026 (UTC)\\[\\] </p>\n<p>My reading of regardless of the outcome was that they would be paid even if they did not meet the publish at least 5 articles successfully requirement. But it's not entirely clear to me, maybe there's another process like you said. I just don't like the idea that someone could spend their time on this and not even get paid for it, because this seems a little bit exploitative to me. --Gurkubondinn (talk) 16:46, 20 January 2026 (UTC)\\[\\] </p>\n<p>Yes, that's what confused me too – reading the job posting, it seemed like they wouldn't be paid for it, which is more worrying. Chaotic Enby (talk · contribs) 17:24, 20 January 2026 (UTC)\\[\\] </p>\n<p>I understand the confusion. The job posting does indeed say 5 articles (which is a mistake, actually it should be saying 2), but this is more a \"protection\" against abuse of ill-intentioned actors, not something that we really have to enforce in practice except from really egregious cases. What we want to avoid is someone signing up, not doing any work at all, and asking for money. If the person has actually tried to onboard but just didn't succeed, and there is evidence of this (e.g., conversations, drafts, etc.) then we've always paid the entirety of the 40h. 7804j (talk) 18:48, 20 January 2026 (UTC)\\[\\]</p>\n<p>Setting asides concerns that this comment was LLM-written or generated in some part, I don't see how holding OKA editors to basic standards of competence and verifiability is holding them to a different standard. The raw volume of errors found and possibly introduced but not yet found is of great concern. We cannot, as a community, keep up with an organized and fast flow of misinformation. This is exacerbated by the pattern of avoiding communication; I was in the 1/1 thread, and it took an outright block to enforce basic communication.</p>\n<p>Also, the general rule in translation is to translate from a foreign language into your native language, with which you are hopefully more familiar. I don't see any indication that this is the case. <strong>I</strong>seulttalk to me 03:10, 20 January 2026 (UTC)\\[\\]</p>\n<p>@7804j, I'd appreciate seeing more details about the onboarding process.</p>\n<ol>\n<li>Do you have an example of a position description or call for expressions of interest from potential new translators? Where do you advertise?</li>\n<li>How are new translators selected and what are the selection criteria?</li>\n<li>What training is provided to new translators, especially around the different policies snd guidelines of the different language Wikipedias? Links to training materials would be especially helpful for this discussion.</li>\n<li>When English Wikipedia editors bring up quality control issues with you, what specific steps do you take to address these with the translator? e.g. do you show them the relevant noticeboard/talk page discussion, do you identify the specific parts of the text that have problems, do you read and discuss the relevant policy together?</li>\n<li>Is anyone other than yourself involved in the selection, training, or supervision of translators?</li>\n</ol>\n<p><strong>ClaudineChionh</strong> (<em>she/her</em> · talk · email · global) 03:49, 20 January 2026 (UTC)\\[\\] </p>\n<p>Translators are selected by other, more senior, translators. I do not personally get involved in the process. This happens through a mix of personal referrals, and postings on Linkedin. We have tested different platforms, and found that it was the most suitable to get a high-volume of diverse profiles without cost. One of our goals is to onboard non-Wikipedians to Wikipedia, so that we grow the total pool of editors, so we intentionally do not recruit directly within Wikipedia. Of course, this comes at the cost of more complex onboarding and more mistakes being done in the beginning, but we see it as part of the value of OKA to train people on the use of Wikipedia. You can find the post template here.</p>\n<p>All our onboarding materials are published on-wiki, such this quick-start guide and the main instructions. We have these additional instructions for non-en Wikipedias. Senior editors conduct the initial resume screening, interview, and also have several video sessions with new editors to walk them through it and answer their questions. During the trial period, their contributions are reviewed in detail by more senior editors, and the degree of review decreases over time as they gain experience. We continue to monitor things such as: spot-checking of articles, AfC rejections, issue tags, mentions of translators into any Wikipedia discussions. We also organize monthly optional round-tables, and have several internal chats where editors can ask questions or share best practices.</p>\n<p>Our core guidelines is that each OKA editor is responsible for directly interacting with the community, and responding to and resolving issues. It's hard to always be aware of every mention of an OKA translator, but whenever this happens we look into it and try to establish the root cause and solutions. This typically results in a mix of changes to our processes or guidelines, reminders to the broader group, and if the translator was at fault, a formal warning (or dismissal for very significant issues or deliberate violations). A second warning typically leads to the termination of the grant. We aim to be as specific as possible, pointing to the specific text or edit, and to the relevant Wikipedia and OKA guidelines or policies.</p>\n<p>Most of these processes are \"self-managed and decentralized\" to the grant-recipients, where senior editors take on these review and management activities. We then have three volunteers (myself and two other people) overseeing the effort, doing an additional degree of spot-checking, designing and building the tools and processes, and acting as escalation layer.</p>\n<p>Generally speaking, we try to be as transparent as possible with the community about the inner workings of OKA, and have been actively seeking suggestions for process improvements. As long as it remains within the spirit of OKA's mission, I welcome any proposed changes to our processes. 7804j (talk) 07:01, 20 January 2026 (UTC)\\[\\]</p>\n<p>In the interest of transparency, would you be willing to identify who else (besides you) has a supervisory or oversight role? At the moment, I am only aware of you as founder, and the lists of translators supported by OKA, which do not indicate who else is involved in supervision and quality control. <strong>ClaudineChionh</strong> (<em>she/her</em> · talk · email · global) 09:22, 20 January 2026 (UTC)\\[\\] </p>\n<p>Additionally, it would be helpful to clarify the responsibilities of the senior translators, regarding selection, management and review of their junior colleagues. As this isn't outlined in the job posting, is there a different position/contract involving these aspects, or is it an expected part of the duties that all translators sign up for? It could be helpful to make this information available on-wiki, especially regarding who is responsible for whom. That way, we know who to ask for help if there is an issue with a newer translator, and OKA's internal review process can work jointly with Wikipedia's. Chaotic Enby (talk · contribs) 09:48, 20 January 2026 (UTC)\\[\\] </p>\n<p>From what I understand – although please correct me if I am wrong – OKA editors are independent contractors, with a very low salary of 397 USD per month for a 40 hour week, which I believe corresponds to 2.29 USD hourly. Given this precarious status, I am worried that more uncertainty in the translator duties may lead to an overloading of responsibilities, which is worrying as independent contractors do not necessarily have the same protections as paid employees. Having more clarity regarding their exact responsibilities and the internal structure through which they work may be necessary to reassure some of us in that regards. Chaotic Enby (talk · contribs) 10:07, 20 January 2026 (UTC)\\[\\] </p>\n<p>There are three of us at the core organization level: myself, @Nickoka (based in Winterthur, Switzerland), and a software developer. Nick handles our internal operations and finances, while the developer is building an open-source stats platform to improve our tracking. Neither are active on-wiki editors, which is why I am the primary point of contact for the community.</p>\n<p>I have updated our list of editors to list the three currently active OKA senior editors who assist with peer-review and onboarding.</p>\n<p>The relationship between senior and junior editors is one of mentorship rather than a strict hierarchy. We don't have rigid \"roles descriptions\" for them; instead, senior editors are long-term grant recipients who have demonstrated a high level of proficiency. They are tasked with:</p>\n<ul>\n<li>Onboarding new grant recipients.</li>\n<li>Conducting peer-reviews during trial periods.</li>\n<li>Coaching on OKA, Wikipedia and local project policies. They are given significant freedom to manage their \"mentee\" groups as they see fit, provided the output meets Wikipedia's quality standards.</li>\n</ul>\n<p>To clarify the stipend amounts: the $397 USD figure is for the initial trial period. For regular editors, this increases up to $452 USD, and for those taking on mentoring/management roles, it ranges between $525 and $634 USD. While these amounts are modest, we ensure they are above the minimum wage in the countries where the editors reside.</p>\n<p>Many of our recipients edit part-time or use the stipend to support their existing volunteer work. Our goal is to enable people to contribute who otherwise wouldn't have the financial flexibility to do so. We are very sensitive to the \"precariousness\" mentioned—which is why we don't treat this as a traditional job with \"overloading responsibilities,\" but as a grant to support independent encyclopedic work. There are also no set working hours—people are free to work whenever they want to, and leave at any time without any notice.</p>\n<p>I am happy to share more details on our internal workflows if it helps. However, our internal documentation includes sensitive payment details, so I need to be careful about where it is shared. I'm open to providing any additional details that the community feels that would help bridge the trust gap. 7804j (talk) 12:40, 20 January 2026 (UTC)\\[\\] </p>\n<p>@7804j While these amounts are modest, we ensure they are above the minimum wage in the countries where the editors reside.</p>\n<p>I have changed my mind. This is disgustingly exploitative. You're a Swiss-based non-profit exploiting residents of the global south to add AI slop to the project and expecting volunteers to clean this up. Take a long, critical, look at yourself and your operations.</p>\n<p>At this juncture I want OKA editors to be <strong>prohibited from directly editing articles entirely</strong>, at the minimum. <strong>qcne</strong> (talk) 16:18, 20 January 2026 (UTC)\\[\\] </p>\n<p>+1 --Gurkubondinn (talk) 16:26, 20 January 2026 (UTC)\\[\\]</p>\n<p>While I don't know in which canton of Switzerland you are based, the lowest minimum wage (in Ticino) is 19.5 CHF, or 24.7 USD. In comparison, the minimum hourly wage for your freelancers is 2.29 USD. Chaotic Enby (talk · contribs) 16:37, 20 January 2026 (UTC)\\[\\] </p>\n<p>I find the characterization of our work as \"AI slop\" and the personal attacks regarding \"exploitation\" to be both disrespectful and inaccurate. I ask that we remain focused on the substance of Wikipedia policy.</p>\n<p>Regarding the stipend levels: OKA is a non-profit providing grants, not a corporation hiring employees. We provide financial support to editors in regions where the cost of living is lower, specifically so they have the freedom to contribute to Wikipedia full-time—something that would otherwise be impossible for them. Comparing Swiss minimum wages to the local economies of our editors is not a helpful metric for a global project. Our goal is to empower contributors from the Global South and under-represented communities by removing the financial barriers to their participation.</p>\n<p>Whether a stipend is \"reasonable\" is a subjective labor discussion that is outside the scope of this noticeboard. The community’s role here is to evaluate the <strong>edits</strong>, not our internal financial structure.</p>\n<p>As for the content, labeling thousands of reviewed translations as \"slop\" ignores the thousands of hours of human work and the value these articles provide to readers.</p>\n<p>I am happy to discuss quality controls, but I will not engage further with personal attacks or hyperbole about our mission. 7804j (talk) 17:05, 20 January 2026 (UTC)\\[\\] </p>\n<p>99% chance AI Generated. We are highly confident this text was AI generated</p>\n<p>It is disrespectful to continue to talk to us via a chatbot, despite already being asked not to. This is now disruptive. I would hat and hide your comment if I was not involved. <strong>qcne</strong> (talk) 17:35, 20 January 2026 (UTC)\\[\\] </p>\n<p>I have been open about using tools to help structure my thoughts and ensure my responses are clear and professional. There is no policy that mandates a specific drafting style, and I find it counterproductive to derail a serious discussion about OKA's operations into a debate over my personal writing process. 7804j (talk) 18:00, 20 January 2026 (UTC)\\[\\] </p>\n<p>Your own use of LLMs, as the founder of the nonprofit in question, to communicate with editors here on Wikipedia, which your nonprofit relies on, is absolutely relevant. --Gurkubondinn (talk) 18:03, 20 January 2026 (UTC)\\[\\]</p>\n<p>Your website indicates that your translators bill you for their time worked; is that because having them act as independent contractors is less legally regulated than if you just granted them the money? I mean, you keep calling them grants anyways… ExtantRotations (talk) 18:06, 20 January 2026 (UTC)\\[\\] </p>\n<p>It's a mix of reasons, but the primary one is process simplicity. We don't need to actively keep track of who's working and who's not working: we wait for the editors to send us the list of what they've worked on, we review it, and pay. Given that we're running this as only 3 volunteers, it's important to keep processes as lean as possible for the volunteers. But this also means when an editor decides to no longer work with OKA, or if they want to reduce their activity rate or not work for a few weeks, they don't even need to let us know, they can just stop submitting their \"bills\". 7804j (talk) 18:51, 20 January 2026 (UTC)\\[\\]</p>\n<p>Apologies but you are the one who created a nonprofit that is financially dependent on Wikipedia. I think it is extremely prudent to ask things like how much the charity raises per year or how much you earn in comparison to your employees. I don’t know about Switzerland but in Canada this info is publicly available for nonprofits. ExtantRotations (talk) 17:40, 20 January 2026 (UTC)\\[\\] </p>\n<p>According to this article from 2025 -- the organization linked to the news article so I assume this is ok to post -- half the budget (80,000 Swiss francs) is from the founder's personal income and half is from \"donations and fundraising.\" 10,000 Swiss francs last year came from a Wikimedia grant to evaluate AI-assisted translations on the encyclopedia (something all of us are currently doing for free, by the way). I don't know whether that 10,000 is part of the \"donations and fundraising\" or separate. Gnomingstuff (talk) 18:15, 20 January 2026 (UTC)\\[\\] </p>\n<p>Actually we have just published the results of this research project here. 7804j (talk) 19:10, 20 January 2026 (UTC)\\[\\]</p>\n<p>From your website (bolding mine):</p>\n<p><strong>To maximize the impact of our limited funds</strong>, we primarily award grants to translators based in countries with low costs of living. These countries also tend to be underrepresented among Wikipedia editors, which also helps us decrease Wikipedia’s systematic bias. ExtantRotations (talk) 17:50, 20 January 2026 (UTC)\\[\\]</p>\n<p>Can you point to an on-wiki discussion where you proposed the translation project to the English Wikipedia community <em>before</em> creating your non-profit association and taking on translators? <strong>ClaudineChionh</strong> (<em>she/her</em> · talk · email · global) 21:59, 20 January 2026 (UTC)\\[\\]</p>\n<p>@7804j: I'd really appreciate a response to this question because as it stands, it looks as if you initiated this translation project without considering the wishes of the broader community or asking for feedback beforehand. <strong>ClaudineChionh</strong> (<em>she/her</em> · talk · email · global) 21:48, 21 January 2026 (UTC)\\[\\] </p>\n<p>As Qcne wrote in the timeline, no, I did not inform the community <em>before</em> starting the project, but many discussions took place shortly after, and at that point I took many steps to act on the community's feedback and be even more transparent and proactively communicating (e.g., formally associating OKA to Wikipedia:WikiProject Intertranswiki). Since then, the community has been overall supportive of our OKA, with the exception of isolated incidents (and this discussion so far, which I believe does not reflect the community's balanced opinion but rather the opinion of a self-selected subset of the community).</p>\n<p>Why did I not inform the community <em>before</em> the project? Simply because at the time, I didn't realize it was expected, as I had very little experience with such types of projects, or even with interacting with the Wikipedia community in general (I was just a regular editor). Also, at the time I wasn't even sure I would be able to make it work, and the concept was still very <em>experimental</em>, thus not having very clear boundaries like it has today 7804j (talk) 07:05, 22 January 2026 (UTC)\\[\\]</p>\n<p>&gt; Can you point to an on-wiki discussion where you proposed the translation </p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mo>.</mo><mo>.</mo></mrow></math></div>\n<em>before</em> <div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mo>.</mo><mo>.</mo></mrow></math></div>?<br>\n&gt; — User:ClaudineChionh<p></p>\n<p>So the answer to her question is a <em>\"no\"</em> then?</p>\n<p>P.S.: please use your own words if you choose to reply to me. --Gurkubondinn (talk) 13:15, 22 January 2026 (UTC)\\[\\] </p>\n<p>Yes the answer is no, which I felt was pretty clear in my previous reply (which, by the way, was <strong>not</strong> LLM-generated). 7804j (talk) 13:47, 22 January 2026 (UTC)\\[\\] </p>\n<p>Apologies, I didn't notice that you had given a clear answer. That's my bad, I wouldn't have written that reply otherwise. --Gurkubondinn (talk) 15:46, 22 January 2026 (UTC)\\[\\]</p>\n<p>Thank you for your response yesterday (which was a long day at work for me). For what it's worth, I disagree with your interpretation that this discussion is not representative of the community's overall sentiments towards your initiative. <strong>ClaudineChionh</strong> (<em>she/her</em> · talk · email · global) 22:09, 22 January 2026 (UTC)\\[\\] </p>\n<p>This discussion involves a dozen of editors that saw it through the admin noticeboard. Over the past few years, hundreds of people have been exposed to OKA's work, and I have personally received many messages of encouragement from a variety of editors and admins. Many editors have also thanked our editors directly for their contributions through barnstars, words of appreciation, etc. Several of our editors are active in Wikipedia projects, some participate in local Wikimedia chapter events, etc. So I don't know what's the overall community's view, but I know for sure that this thread doesn't represent it well 7804j (talk) 06:26, 23 January 2026 (UTC)\\[\\] </p>\n<p>I second that. People don't come to AN to heap praise and give pats on the back, this is the place for grilling and hard knocks :P Piotr Konieczny aka Prokonsul Piotrus| reply here 06:28, 23 January 2026 (UTC)\\[\\]</p>\n<p>This is ridiculous OKA did not come up with outsourcing and the fact that labor has different cost in different places. We live in the real world. WP has a massive lack of representation from the global south. Many in the global south do not have the luxury to spend their unpaid time on the internet. OKA is not forcing them to edit wikipedia. Some of these editors may well see this as a great opportunity for them to contribute to a project. It may be a much better opportunity for making money than other available to them. If they want to argue OKA is doing unfair labor practices then let's hear them say it, but it not for you to decide for them. Czarking0 (talk) 04:07, 11 February 2026 (UTC)\\[\\]</p>\n<p>At this juncture I want OKA editors to be prohibited from directly editing articles entirely, at the minimum. I agree in full. Stepwise Continuous Dysfunction (talk) 19:03, 21 January 2026 (UTC)\\[\\] </p>\n<p>Firmly agree. There are numerous issues here - practical issues, labor issues, and community interaction issues - of which the most significant is that users are being told to add translations that they are not able to check for errors and copyedit. 7804j's responses here have made it clear they cleared this nonprofit without considering what the community would want, and have no intention of addressing the numerous major concerns. This effort needs to be shut down, and not allowed to simply skip over to harming a different project. I would prefer if someone else can formally propose the prohibition - I don't have any experience in doing so - but I will be it myself if need be. Pi.1415926535 (talk) 21:23, 21 January 2026 (UTC)\\[\\]</p>\n<p><strong>I could not agree more</strong>, the responses here by 7804j are concerning, almost all are AI generated and they seem to not respect the community's wishes at all. <em>– LuniZunie</em>(talk) 23:21, 21 January 2026 (UTC)\\[\\] </p>\n<p>I beg to differ. We don't know what community wishes are, except in the broad context of building an encyclopedia, and 7804j and his project have been certainly doing just that. As for the use of AI to streamline replies, there's a difference between having an AI reply, and having an AI rewrite/copyedit one's thoughts, and my understanding is that they've been doing the latter, not the former. While we are seeing a norm clash, with some folks considering the latter lazy, I think it's a sign of things to come, and overall an improvement, given low quality of online discourse; in either way, while I also support banning \"using AI to talk\", I have no problem with \"using AI to help one refine one's response\", and per WP:AGF I see no reason to believe this is not the case here. Lastly, I am also open to the consideration that AI may help some folks suffering from forms of disability (dyslexia, etc.) and I find criticism of folks that may benefit from such tool unwarranted and arguably, distasteful. Would you criticize a blind person for using a gadget to edit Wikipedia? In the end, it doesn't matter what tool one uses to write a response and communicate, as long as we are still talking to them (and not a bot). Ban the bots, let real people do what they want - it's not our place to force our thoughts on which tools they should be using. Piotr Konieczny aka Prokonsul Piotrus| reply here 05:25, 22 January 2026 (UTC)\\[\\] </p>\n<p>@Piotrus I absolutely agree with your point on the disabilities as I have dysgraphia and dyslexia, and I do use AI to help with spelling and reading (because I can't read for the life of me). However, what I am trying to say is, if an issue is brough up with you, and you use a bot to respond, that's a problem; you aren't responding or acknowledging the issue at all by doing that.As per the community's wishes, that is true that we only know the broad context, but what I meant by that is that the responses here, and the subsequent responses by 7804j, are showing there is a pretty large mismatch in what is wanted. In a perfect world, we would tell 7804j that everything needs to be properly checked, but this isn't a perfect world and that just won't happen and I am not given confidence by the responses here to prove otherwise.I of course, agree that minimal AI usage should be fine, I mean, it's the future whether we like it or not, and failing to adapt to the times will be this project's demise. But I just don't think this is the way to do it. <em>– LuniZunie</em>(talk) 12:25, 22 January 2026 (UTC)\\[\\] </p>\n<p>Well, I have read chunks of the discussion here, maybe I missed something. It's clear some OKA translators got lazy and used AI without double checking for common errors. I expect OKA (=7804j) to introduce safeguards to prevent this from happening, and discipline repeated offenders (because as everyone knows, if we have to do it here, it won't be pretty). I expect OKA will agree - it's not their first rodeo, and in the past, AFAIK, they have been ammenable to improve their standards and ensure the content they deliver is acceptable. Piotr Konieczny aka Prokonsul Piotrus| reply here 12:51, 22 January 2026 (UTC)\\[\\]</p>\n<p>I am a bit confused as to what gives the impression that OKA isn't willing to adapt. As Piotr mentioned, we are more than happy to make any changes to our processes that the community finds beneficial.</p>\n<p>Also, as Piotr mentioned, I only used AI to organize my thoughts, I did not use AI to generate messages from scratch. It is therefore fully within the community's own guidelines. Also I have been extremely detailed and transparent in my responses.</p>\n<p>LuniZuni, is there a concrete ask that you feel OKA should commit to but where you feel there is a fundamental mismatch between what you would want, and I would be willing to agree to? 7804j (talk) 13:39, 22 January 2026 (UTC)\\[\\] </p>\n<p>LuniZuni is my doppelgänger :pis there a concrete ask that you feel OKA should commit to but where you feel there is a fundamental mismatch between what you would want, and I would be willing to agree to?<br>\nI'm not asking about what I want, what I want is a website where AI is allowed in moderation, but not in content generation. What I am asking is that these things be communicated onwiki before they occur, not in discussions that take place after the fact, they should be done before. This is a major thing being done by your non-profit, that needs to involve the community. Right now, it seems like the community is involved, but only after the fact, which is not very helpful. <em>– LuniZunie</em>(talk) 13:47, 22 January 2026 (UTC)\\[\\] </p>\n<p>In all honesty, I (and others) complain elsewhere about how WMF does stuff without consulting community. Of course, we seem to be unable to change WMF behemoth; here, we have much more leverage, but I also think - again, based on past experience - we are much more likely to see OKA change what they do than WMF, sigh. Piotr Konieczny aka Prokonsul Piotrus| reply here 14:51, 22 January 2026 (UTC)\\[\\] </p>\n<p>While I of course agree, I don't think that just because WMF does it means the rest of us should also do it. <em>– LuniZunie</em>(talk) 14:53, 22 January 2026 (UTC)\\[\\]</p>\n<p>I think the WMF comparison is very salient here. OKA clearly has money and wants to support the project. Their vision is not 100% aligned with all the community members. Neither is WMF's. However the community may be able to play OKA and WMF off against each other for our own gain. Though I doubt we have the skill to do it. Czarking0 (talk) 04:11, 11 February 2026 (UTC)\\[\\]</p>\n<p>Well I don't agree but their is no point in just saying you agree or not. Czarking0 (talk) 04:08, 11 February 2026 (UTC)\\[\\]</p>\n<p>I am concerned about the accusation of exploitation made here; this is serious and I worry it was not fully considered in the heat of the moment. 7804j has said OKA work is not meant to be full-time, but fine, let's say people do it full-time anyways. 7804j says workers are paid more than minimum wage, but that doesn't mean much; even in Switzerland, most of the country has no minimum wage, so he could be paying one Rappen an hour and that would still be true. But he has given us numbers here that we can check against the statistics. Let's look at two countries in the global south with vast numbers of native English speakers: Nigeria and India. The World Bank says average annual income in Nigeria was 1663 USD in 2021 </p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mn>1</mn></mrow></math></div>, which comes out to 139 USD per month. For India it gives 1907 USD per year, or $159 per month. The same list gives $115 monthly for Senegal, a (somewhat) Francophone country, and $427 monthly for Colombia, a Spanish-speaking country. (Many OKA articles are translated from Spanish or French. These are the first countries that came to mind for each language – if I were to cherry-pick the data, I could list, say, Chad and Bolivia instead.) For comparison, the minimum OKA stipend listed above was $397. Now, I don't know if the World Bank data truly represent a living wage in those countries, something that is very difficult to measure, and I don't know where OKA editors are (maybe they're in countries with far higher cost-of-living?). But this is a reasonable estimate, and it would seem to contradict the accusations of exploitation made above. <strong>Toadspike <div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>T</mi><mi>a</mi><mi>l</mi><mi>k</mi></mrow></math></div></strong> 04:22, 24 January 2026 (UTC)\\[\\] <p></p>\n<p>Thank you for looking into this, I was also rather shocked by the language used. No good deed, you know. Here, someone is paying out of their own pocket (or at least partially, and in either case not counting time spend managing this) to get volunteers (nobody is forcing anyone to do anything) to contribute to Wikipedia - and he is being accused of exploitation. I am at a loss of words, really, because I don't think anything I could say would be particularly nice. Sigh. Piotr Konieczny aka Prokonsul Piotrus| reply here 04:33, 24 January 2026 (UTC)\\[\\]</p>\n<p>Agreed, I left a different comment with the same idea above Czarking0 (talk) 04:12, 11 February 2026 (UTC)\\[\\]</p>\n<p>@7804j: This sub-thread got somewhat derailed. I want to return to the issue of OKA's labour force. Set aside dollar/euro/franc amounts for a moment. To state my bias upfront, as an Asian immigrant to Australia I am very sensitive to \"representation\" being used as a fig leaf for exploiting workers in the global south, and \"independent contractor\" arrangements used to absolve your responsibility to your workers. Because that's what they are, <em>your workers</em>, and treating their half- or full-time employment as gig work is abhorrent to me. You have imposed a mass article creation program on the English Wikipedia without prior community discussion, which means the existence of this program and your workers' livelihoods is dependent on the community's goodwill. A noticeable number of your workers were given inadequate training or oversight which meant that they came to the community's attention as \"problem\" editors, though it's now becoming clearer to me that these problems are due to systemic issues within OKA itself. You are squandering any remaining goodwill with your evasive answers here. Like qcne I have become disgusted with this whole situation, but am also sensitive to the needs of the OKA workers who presumably have no form of protection from the consequences of their boss's behaviour. <strong>ClaudineChionh</strong> (<em>she/her</em> · talk · email · global) 04:35, 24 January 2026 (UTC)\\[\\] </p>\n<p>And I look at it from exactly the other side. The boss is trying to help, both the workers and the community, and should not be held responsible for the fact that some of his workers are lazy and cut corners. The boss is volunteering and using their own money, not making a dime of profit. The workers are the ones who are making profit - and nobody is forcing them, they are free to choose other gigs. I can't understand how come we are criticizing a volunteer who invested immense amount of time, and their own financial resources, to help our community, instead of the paid editors, some of whom are obviously abusing the system. Piotr Konieczny aka Prokonsul Piotrus| reply here 14:54, 24 January 2026 (UTC)\\[\\] </p>\n<p>Who else's problem <em>but</em> the boss' can it be if the translators are not adequately trained and their work is not adequately reviewed? It's massively irresponsible to sic a cadre of unsupervised 'translators' on Wikipedia with the specific instruction to use general-purpose LLMs to translate massive swathes of text and not implement any kind of oversight or verification. OKA can't then wash their hands of that and say \"well, some people are just lazy and don't review their work\". Of course they are, that's why you have to plan for that in your operational model! For instance, they could have one or two more highly paid staff whose sole job is to internally review the translations made by the 'grunt' translators before they go live. Or they could at least implement some kind of peer-check system between the translators themselves. Either way, if you just leave them to their devices, <em>obviously</em> you're going to end up with one or two of them just shovelling slop into the encyclopedia with no second thought. Athanelar (talk) 10:58, 25 January 2026 (UTC)\\[\\] </p>\n<p>Although I disagree with your assessment of OKA, which I think is quite responsible, I do think there is merit in the idea that they should hire some folks to do quality check, even if this would reduce the total output due to reducing the translator's total number a bit. It would, hopefully, prevent the issues we are seeing (as marginal as they are, IMHO). @7804j - this idea is worth considering seriously. Piotr Konieczny aka Prokonsul Piotrus| reply here 14:18, 25 January 2026 (UTC)\\[\\]</p>\n<p>I agree that having peer reviewers could make sense, and I've started to think about it in the last couple of days. However, it's not easy to put in place operationally and there are several different options to consider, so it will require some further thinking and fleshing out.</p>\n<p>What I can do is draft a process and start a pilot internally with a few editors and focused on EN WP. If this works well, we can then scale it to the rest of OKA and other Wikipedias. This might take a bit of time, though I expect we can already get some results and feedback within a few weeks.</p>\n<p>-----</p>\n<p><em>As a side note, I really dislike the \"boss\" analogy. Would we require that every Wikipedia editor has a boss and an assigned reviewer? The whole point of OKA is to provide grants so that people can become regular editors. The more process we put on them, the more they lose the freedom that makes Wikipedia great in the first place.</em></p>\n<p><em>OKA doesn't strictly <strong>require</strong> that grant recipients translate using LLMs, nor that they translate in fact. These are mere recommendations based on what we have found to be an easy way for people with dual language proficiency to have significant impact on Wikipedia. We do actually encourage editors to explore any other tools or types of Wikipedia contributions, but relatively few do, simply because they agree that LLM translations are a good way to have positive impact.</em></p>\n<p><em>One of OKA's goal is actually to bring new editors to Wikipedia, which is why we don't advertise <strong>on</strong> Wikipedia. If someone promotes Wikipedia at their university or workplace, and as a result some mix of editors start editing, some being good and some being less so, is the editor who promoted Wikipedia in the first place going to be held accountable for having promoted it to the wrong people?</em></p>\n<p><em>That being said, I acknowledge that due to the scale and organized nature of OKA, additional safeguards are useful, even if they restrict a little bit the freedom of OKA editors, hence my willingness to try the peer reviewer approach.</em> 7804j (talk) 14:38, 25 January 2026 (UTC)\\[\\] </p>\n<p>Your goal is understandable and admirable, but you of course have to be aware that paying people for their contributions is going to change the way they contribute. Similarly, of course someone who tells their friends \"hey, editing Wikipedia is fun, you should try it\" isn't liable for what those people do, but that's not what you're doing. You're specifically selecting, enrolling and supporting paid editors. You <em>are</em> responsible for the people on your payroll, there's no getting around that, because the mere fact that you are paying them is an explicit endorsement of their editing activities.</p>\n<p>The \"one or two bad apples\" argument only works if you're making an active effort to catch the bad apples. It's very good that you take action when it's reported to you, but it's concerning that somebody is able to go on for so long without it ever being caught internally. You <em>need</em> to have some kind of internal review process, otherwise you can't be surprised when people blame you for the actions of people you're literally paying to edit. Athanelar (talk) 14:49, 25 January 2026 (UTC)\\[\\] </p>\n<p>I don't agree with this perspective. If a university gives a full scholarship to a student (tuition and costs of living), it doesn't automatically become liable to all the activities and publications of that student, even if the university encourages its students to write and publish. E.g., if the student publishes a flawed and questionable paper on their own blog, the university may decide to cancel the scholarship, but shouldn't be held liable for having granted that scholarship in the first place. 7804j (talk) 16:22, 25 January 2026 (UTC)\\[\\] </p>\n<p>Sure, but if one of the students accepted for said scholarship demonstrated that they lacked the qualifications one would expect for a recipient of that scholarship, it would raise questions as to how thoroughly the university is vetting its applicants. Athanelar (talk) 16:50, 25 January 2026 (UTC)\\[\\] </p>\n<p>I disagree with this as well. I know many universities or systems where scholarships are funded without particular vetting of the applicant's competence, other than on the basis of a CV.</p>\n<p>In either case, nobody would say that such universities are <em>the boss</em> of the scholarship recipients, even if such scholarships are tied to certain outcomes (e.g., successfully meeting course criteria). This is the model we're trying to replicate with OKA 7804j (talk) 19:28, 25 January 2026 (UTC)\\[\\]</p>\n<p>@7804j, a brief comment on the \"boss\" wording: I didn't use it in relation to the editing workflow, but to draw out the economic and power dynamics in OKA's structure, regardless of whatever legal or taxation arrangement is used. @Athanelar has picked up on that aspect with paying people for their contributions is going to change the way they contribute. On reflection I think my critique of your gig work arrangement is a tangent from this specific discussion of quality control, but I am still extremely concerned (indeed, still disgusted) by this arrangement and may seek a better venue to continue that discussion. <strong>ClaudineChionh</strong> (<em>she/her</em> · talk · email · global) 00:11, 26 January 2026 (UTC)\\[\\] </p>\n<p>Do keep in mind WP:BOOMERANG. Piotr Konieczny aka Prokonsul Piotrus| reply here 01:22, 26 January 2026 (UTC)\\[\\]</p>\n<p>Apologies for barging in, I'm mostly a lurker, but reading <strong>that</strong> amount of money the organization offers as \"payment\" is decently less than minimum wage in about half of latam, this is straight up exploitation.</p>\n<p>Adding my two cents, OKA should straight up be banned from wiki; AIslop without signs of actually wanting to improve the jank it spits out, and, exploitation of people in need, worse of all, a Swiss non-profit, straight up a neo-colonialist scheme.</p>\n<p>/rant, this kinda things make me angry \\~2026-43687-3 (talk) 19:53, 6 February 2026 (UTC)\\[\\] </p>\n<p>And comments like above make me angry (so what?). Misrepresenting things (there's plenty of evidence OKA is trying to improve), and calling online gig that nobody is forcing anyone to take \"exploitation\", sigh. You think OKA is offering too little? Offer more, yourself. They are at least trying. Piotr Konieczny aka Prokonsul Piotrus| reply here 01:26, 7 February 2026 (UTC)\\[\\]</p>\n<p>There's also the Grok angle: that is the worst LLM one could possibly be using from an \"oh this will be <em>bad</em> bad if it gets out\" perspective, given that the three things the average person associates Grok with nowadays are Elon Musk, revenge porn, and Grokipedia.</p>\n<p>I don't buy the \"Grok knows wikitext\" angle, all modern LLMs have a rudimentary-enough knowledge of wikitext to match the rudimentary level of the translations. Gnomingstuff (talk) 17:45, 20 January 2026 (UTC)\\[\\] </p>\n<p>Wikipedia was mostly written by volunteers. But sometimes they need a bit of help helpful, indeed. <strong>qcne</strong> (talk) 17:52, 20 January 2026 (UTC)\\[\\]</p>\n<p>We have actually just published the results of a study we have done on LLM quality for Wikipedia translation. Based on these results, we have proceeded to switch over from Grok to Claude and ChatGPT as the recommended tools, but there are certain use cases where Grok still performs better.</p>\n<p>We had done a similar (though a lot less detailed) analysis in the past, and so far Grok was always outperforming the other models for this specific task. So we haven't picked Grok because we like Elon Musk or Grokipedia (in fact, I don't), but just based on data. 7804j (talk) 19:15, 20 January 2026 (UTC)\\[\\]</p>\n<p>The mere willingness to <em>test</em> using the revenge-porn generator, the one that declared itself \"MechaHitler\", for this purpose is indicative of poor judgment. Stepwise Continuous Dysfunction (talk) 19:05, 21 January 2026 (UTC)\\[\\] </p>\n<p>And CSAM generator. The level of doublethink needed to convince oneself of acting for the public good while wanting to be anywhere near that is breathtaking. — <strong>Hex</strong> • <em>talk</em> 13:35, 22 January 2026 (UTC)\\[\\] </p>\n<p>I find this response and the one above extremely concerning. Are you suggesting that because Grok is a very controversial tool that sometimes produces very problematic results when prompted to do so, that even the <em>idea</em> that I would want to <em>test</em> the effectiveness of Grok as a <strong>translation</strong> tool is a proof that we have poor judgement or poor intents?</p>\n<p>Keep in mind that translation is a very mechanical task. Grok isn't going to suddenly add Hitler references or generate child porn when asked to translate a sentence about the History of France... I'm really not a fan of Elon Musk, but come on, there's really no need for extra sensationalism here... 7804j (talk) 13:43, 22 January 2026 (UTC)\\[\\] </p>\n<p>Are you suggesting that because Grok is a very controversial tool that sometimes produces very problematic results when prompted to do so, that even the idea that I would want to test the effectiveness of Grok as a translation tool is a proof that we have poor judgement or poor intents? yes. <strong>qcne</strong> (talk) 13:49, 22 January 2026 (UTC)\\[\\] </p>\n<p>This is ridiculous. 25 years ago I could have said the internet is a disgusting tool for sharing CSAM and you want to test using it to distribute an encyclopedia!? Czarking0 (talk) 04:22, 11 February 2026 (UTC)\\[\\]</p>\n<p>If you knew a human who created CSAM, would you still work with them and trust them? <em>– LuniZunie</em>(talk) 13:51, 22 January 2026 (UTC)\\[\\] </p>\n<p>And now for argumentum ad hitlerum... Piotr Konieczny aka Prokonsul Piotrus| reply here 14:52, 22 January 2026 (UTC)\\[\\] </p>\n<p>Unless I am misunderstanding something, I don't see how that applies. <em>– LuniZunie</em>(talk) 14:55, 22 January 2026 (UTC)\\[\\]</p>\n<p>&gt; Keep in mind that translation is a very mechanical task.</p>\n<p>I believe that this shows a rather poor understanding of what goes into localisation and translation (and the unwarranted editorialising that LLMs perform whenever they're asked to <em>merely translate</em> something). --Gurkubondinn (talk) 08:17, 23 January 2026 (UTC)\\[\\] </p>\n<p>I am not saying it is \"mechanical\" in absolute terms, but compared to other prompts you might give an LLM like \"Tell me what you think of Nazis\" 7804j (talk) 11:38, 23 January 2026 (UTC)\\[\\] </p>\n<p>Ah, so you meant \"mechanical task\" in the context of LLM inputs/prompts? --Gurkubondinn (talk) 11:51, 23 January 2026 (UTC)\\[\\] </p>\n<p>Yes 7804j (talk) 11:56, 23 January 2026 (UTC)\\[\\]</p>\n<table>\n<thead>\n<tr>\n<th>Aside. <strong>I</strong>seulttalk to me 01:33, 23 January 2026 (UTC)\\[\\]</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>Also, fun fact. Wikipedia was created by one Jimbo Wales. From his bio: he also founded (earlier) \"a web portal featuring... softcore pornography\". How does it make you feel? :P Piotr Konieczny aka Prokonsul Piotrus| reply here 14:54, 22 January 2026 (UTC)\\[\\] Well, I think CSAM and softcore pornography are a bit different... <em>– LuniZunie</em>(talk) 14:57, 22 January 2026 (UTC)\\[\\] Sure, just as Grok and CSAM are a bit different... Piotr Konieczny aka Prokonsul Piotrus</td>\n</tr>\n</tbody>\n</table>\n<p>With all due respect, I think that the point the above users are trying to make is not that Grok is likely to do any of these things while translating Wikipedia articles, but rather, that using a model that is particularly notorious for its poor oversight was a lapse in judgement. While use of Grok in this context is not necessarily an endorsement of its creator or its controversies, even in the absence of your explicit dis-endorsement above, it's not unreasonable that people aware of the many controversies of Grok would be concerned about its use in this context. It's obvious that using Grok for translation purposes is not likely to produce questionable or illegal image content, or that OKA editors would be foolish enough to put such content on Wikipedia, however, Grok's biases are well documented, and its tendency to gas up or disparage particular concepts and individuals based on the politically-motivated instructions provided by its creators is relevant in the context of translating encyclopedic text. If paired with poor oversight, it's reasonable to be concerned that text that is unencyclopedic or outright biased may be more likely to slip through when using Grok as opposed to other models with less toxic reputations. It's good to know that you've switched over to primarily relying on Claude and ChatGPT (despite having their own problems, I think most users would consider them far less likely to produce biased outputs), but I hope you can understand why so many users are concerned and consider the use of Grok to have been a poor decision, even though I'm sure it was done with good intentions. Ethmostigmus 🌿 (talk | contribs) 02:51, 23 January 2026 (UTC)\\[\\] </p>\n<p>I understand, and I have definitely taken these concerns into account. However my own extensive experience of using Grok was that it's only showing these biases or issues when either (a) prompted to do so, (b) on extremely rare occasions (that I never personally observed other than news report) where prompted about specific topics that are very close to Elon Musk. As you mentioned, for translations that would never happen due to the mechanical nature of the task.</p>\n<p>So I am not surprised that some people in this thread would <em>think</em> that it's a sign of poor judgement, but the reality is that it isn't. It was carefully considered, and at the time, Grok indeed had a much better performance than other models, so the pros significantly outweighed the cons back then. Fortunately we can now switch to models that don't have this reputation since the data proves they now also perform better 7804j (talk) 06:36, 23 January 2026 (UTC)\\[\\]</p>\n<p>Context: This section discusses concerns regarding OKA's use of LLMs for translation, including issues with edit quality, overwrites, coordinated mass-editing, and quality assurance. It also details existing discussions and reports related to these issues, as well as suggested remedies.</p>\n<hr>\n<p>Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] </p>\n<p>To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] </p>\n<p>And that would be just fake news and should be ignored. AI translations that are vastly accurate (reported problems account for only a small minority of generated content) is not a \"flood of AI content\". Piotr Konieczny aka Prokonsul Piotrus| reply here 04:07, 23 January 2026 (UTC)\\[\\] </p>\n<p>OK, but what about unreported problems? No one's gone and systematically checked all of them; the above dataset seems to mostly look for spelling error-type issues, whereas the issues AI Cleanup has found are far deeper. Most of <em>everything</em> the project has found, really, was undetected for months or years. Gnomingstuff (talk) 08:46, 23 January 2026 (UTC)\\[\\] </p>\n<p>If they are unreported, we cannot even be sure they are real. Maybe there are no problems to find. All that we have is a proverbial storm in a teacup - one OKA translator was lazy and did not check the AI output. \"Big deal\". IF there were more translators like that - I haven't seen the diffs or usernames. Piotr Konieczny aka Prokonsul Piotrus| reply here 11:35, 23 January 2026 (UTC)\\[\\] </p>\n<p>To be clear, given we had over 50 translators that each published dozens or hundreds of articles, I am quite sure we will find more examples if we really look for them. The question for me isn't \"do problems happen\", but \"how often do these problems happen, do we believe we can mitigate them, and if not are they so large that they offset all the great outcomes where no problems were found\" 7804j (talk) 11:42, 23 January 2026 (UTC)\\[\\] </p>\n<p>Indeed. To answer the first and third questions, we need data I am not sure we can easily get (and without the data, it's all pretty much subjective ILIKEIT/IDONTLIKEIT, sigh). To answer the second - well, that's something you can do (and perhaps have done already here?). Piotr Konieczny aka Prokonsul Piotrus| reply here 11:51, 23 January 2026 (UTC)\\[\\] </p>\n<p>Yes, the second point is something that we've tried to do as much as we could, but where we're fully open to any suggestions for doing things differently 7804j (talk) 11:57, 23 January 2026 (UTC)\\[\\]</p>\n<p>From what I've seen in the past, OKA delivers good quality content, although as with any large project, it will occasionally fail and produce a dud. I was not aware OKA has adapted LLM, but it makes sense - these are popular and efficient tools (and, of course, imprefect). The only realistic suggestion I've is that all OKA content output should be checked with tools like ChatZero, and that the translators are required to ensure that LLMs didn't add hallucinations, or lazily skipped over content. Other than that - good job, kudos to volunteers engaged in it, and move on. (And, sigh, it amazes me how much complains and scorn is heaped upon volunteers who go an extra mile and try to something ambitious). --Piotr Konieczny aka Prokonsul Piotrus| reply here 05:11, 22 January 2026 (UTC)\\[\\] </p>\n<h4>Arbitrary break</h4>\n<p>\n</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>e</mi><mi>d</mi><mi>i</mi><mi>t</mi></mrow></math></div>\n<p></p>\n<p>@7804j, I'm not an admin but I've got a few questions:</p>\n<ol>\n<li>I'm wondering how general editors would know to approach you if there were any concerns over an editor? The user-page disclaimers go to your website, but the Contact page of your website seems to be down and there's no indication of who we should contact on Wikipedia. Are you expecting general editors to report them using our normal escalation process then admins have to contact you in turn?</li>\n<li>Do you perform regular quality checks (e.g. checking Talk pages for warnings) to ensure that quality is maintained, considering they're being paid? From what I've read above, it looks like they're only checked during probation then left to continue on their own (with self-reporting) but I may well have missed something.</li>\n<li>What happens if an editor has passed probation but their work falls below the expected standard? Has this happened yet?</li>\n<li>I'm also wondering how you accommodate for different projects having different guidelines/policies (e.g Wikipedia:Notability), considering you're essentially transplanting articles from one Wikipedia into a completely different one. I see there's a preset list of articles, but you also advise editors that they're free to create their own.</li>\n</ol>\n<p>Apologies if some of these have been answered, but this is a long thread! Blue Sonnet (talk) 04:53, 23 January 2026 (UTC)\\[\\]</p>\n<ol>\n<li>The OKA.wiki contact page is a redirect to mailto:info@oka.wiki so it seems to work for me. Regarding how to contact us: every OKA editor has a link in their talk page to our meta Wiki profile. There, people can find the username of senior editors and of myself, the website of OKA, and other information. People have used various channels to contact us, including tagging me, writing in my talk page, writing in one of the OKA's talk pages (which are monitored), or emailing us, etc. We don't have a particular preference or expectation on how people reach us -- we expect that most issues get resolved directly by communicating with each editor, and that if people want to escalate an issue, they get in touch with me or with OKA through one of these channels. If you feel this could help, we could for example try to make it more prominent on the Meta page what's the process to escalate issues?</li>\n<li>Yes -- we have a script that lists all the maintenance tags of all pages ever created by an OKA editor (based on the OKA template added in the talk page). The senior editors regularly go through the list to ensure these issues are resolved, so even if a page gets tagged after 1-2 year, we still go back to it to fix the issues.</li>\n<li>Yes this has happened, including for several of the issues listed in this thread. This was discussed in some of the above comments, but we then try to look into what happened, identify if these were process issues or individual issues, and if it's something that can be \"resolved\" or a fundamental skill or attitude problem with the editor. Typically, if this is the first time, it leads to a formal warning, but has led on several occasions to a suspension of the grant to the editor when the issue occurred again. In such cases, we then task the senior editors to review all the work of that editor to see if some \"clean up\" is required to fix any potential issues introduced</li>\n<li>Generally we tell people to follow the English Wikipedia as minimum standard, even when translating to other Wikipedias. But we do have some language-specific guidelines as well to complement these. Editors are indeed free to create their own articles, but in my experience this almost never happened, because people prefer to translate, which feels safer or more familiar for most as it's their profession.</li>\n</ol>\n<p>7804j (talk) 20:04, 23 January 2026 (UTC)\\[\\] </p>\n<p>@7804j Sorry to ask yet another question, but is there a list somewhere of the targeted donations, and/or of which articles fall under them? I couldn't find it, although I believe it probably must be available somewhere. It could be reassuring to make it prominently visible so admins don't worry about issues of paid editing disclosure. Chaotic Enby (talk · contribs) 23:46, 23 January 2026 (UTC)\\[\\] </p>\n<p>Currently not, because so far we haven't received any targeted donations. This statement on OKA's website is more about keeping that option open for the future. If we ever do, yes, I would list them publicly somewhere, with the details of the rules/requirements and affected articles (note that the affected articles could be vaguely defined, e.g., \"Any article that is related to Switzerland\" if the donation was from the Swiss government). 7804j (talk) 07:37, 24 January 2026 (UTC)\\[\\] </p>\n<p>Thanks a lot, that is indeed very reassuring to know! Chaotic Enby (talk · contribs) 07:47, 24 January 2026 (UTC)\\[\\]</p>\n<p>Thanks for the clarification - I'm on mobile so perhaps something went wrong there? It would be nice to have a link to the meta page on the disclaimer, I'd find that really helpful for those who aren't familiar with the project &amp; want to find out more on-wiki. Blue Sonnet (talk) 07:44, 24 January 2026 (UTC)\\[\\]</p>\n<p>I think that we've gotten a little sidetracked above with discourse about specific LLMs and LLMCOMM. My concerns, as expressed earlier, remain unaddressed. The last three discussions originally linked by qcne are troubling. The fact that obvious indicators of synthesis (at best) or hallucination are in translations is unacceptable and has always been so on this project. Less obvious/more insidious errors require more-or-less fully bilingual editors to identify and redress. That more errors haven't been found in the array of huge edits (</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mn>3</mn></mrow></math></div>) made as a pattern from OKA-affiliated editors is almost certainly a question of volunteer scale and capability. AINB is already dealing with a lot. I do not want this community to be in a position where ten unpaid volunteers will have to verify the output of thirty paid editors while hamstrung. <strong>I</strong>seulttalk to me 08:27, 23 January 2026 (UTC)\\[\\] <p></p>\n<p>I am not convinced. Until someone finds more errors, it's pretty much fearmongering/ABF, particularly given the tremendous positive contributions made by OKA (hundreds of translated articles, if not thousands, with only a fraction ever being a problem). We should focus on real, not imagined, problems. Piotr Konieczny aka Prokonsul Piotrus| reply here 11:40, 23 January 2026 (UTC)\\[\\] </p>\n<p>Thanks you. You've made your position abundantly clear in many comments above: AI translations good, no sanctions needed, close this thread. What this amounts to is throwing the burden of verifying the translation on other people and dismissing their concerns. You and 7804 have still not addressed the concerns and diffs originally raised in this thread. You have not provided proof that the translations and expansions are actually good.</p>\n<p>I spent an hour trying to verify four sources here a few weeks ago. It turns out that AI has a really high rate of making things up. I assumed good faith, just as I am here, and got burned. Why don't you go verify everything in this edit, for example? Or, if some error crops up in an OKA editor's contributions, will the blame be attributed solely to that editor and not systemically? It turns out that verification isn't easy!</p>\n<p>Making things up is manifestly forbidden! Why is it that I have never seen an AI proponent involved with AINB or patrolling NPP for AI spam? <strong>I</strong>seulttalk to me 18:16, 23 January 2026 (UTC)\\[\\] </p>\n<p>Why is it that I have never seen an AI proponent involved with AINB or patrolling NPP for AI spam? That would involve work, and the point of AI is to offload work and not care about the results. Cremastra (talk <strong>·</strong> contribs) 18:52, 23 January 2026 (UTC)\\[\\]</p>\n<p>I am an AI proponent that patrols NPP and AfC. Though last month I quit the NPP backlog drive over insufficient use of bots in the queue maintenance. The answer to your question is you have confirmation bias and you do not look to disprove yourself. Czarking0 (talk) 04:26, 11 February 2026 (UTC)\\[\\]</p>\n<p>Noting that, after previous issues were brought up, I decided to check the (at the time) latest OKA draft by that same editor, and found that the very first source I could verify was hallucinated. Of course, every find can be dismissed as \"only a fraction\" given the amount of text OKA publishes, but when this is the case for every translation that gets closely reviewed, statistically, this doesn't bode well. Chaotic Enby (talk · contribs) 21:00, 23 January 2026 (UTC)\\[\\] </p>\n<p>It would help to have a full list of articles to audit rather than finding them piecemeal. Gnomingstuff (talk) 21:25, 23 January 2026 (UTC)\\[\\]</p>\n<h3>A proposal</h3>\n<p>\n</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>e</mi><mi>d</mi><mi>i</mi><mi>t</mi></mrow></math></div>\n<p></p>\n<p>Alright. I don't think we're going to get anywhere particularly helpful out of the forgoing discussion. I would like to remind everyone that we have two core ideals to inform ourselves of what to do here, namely WP:V and \"anyone can edit\". That is, we care very deeply about whether things are verifiable, but we also care very deeply about how this is an encyclopedia written by amateurs working together; that is, mistakes are part of the process, and we can expect new editors in particular to make a lot of them. I agree that OKA probably has something like \"the WikiEdu problem\", namely that people mostly only notice when it's going <em>wrong</em>, not when things are working, which happens far more often. And it also seems to me that OKA editors are being treated as a monolith, so we end up with \"OKA screwed up again\" not \"editor x screwed up again\". Neither of these are fair. But the problems are real. So let's set some rules for OKA translators:</p>\n<ol>\n<li>everything an OKA translator adds to en-wiki must be <em>verified</em> by the translator; that is, if they haven't checked the information against the provided source, it should not be added.</li>\n<li>OKA translators who have received, within six months, four (correctly applied) warnings about content that fails verification will be blocked without further warning if another example is found. (clarification added asilvering (talk) 04:43, 24 January 2026 (UTC))\\[\\]</li>\n<li>content added by an OKA translator who is subsequently blocked for failing verification may be presumptively deleted as WP:V failures unless an editor in good standing is willing to take responsibility for it.</li>\n<li>effective henceforth; that is, no blocks/pdels for content written before this proposal passes (if it does).</li>\n</ol>\n<p>The \"within six months\" clause is there to prevent a scenario whereby a new OKA editor messes up three times and has the sword of Damocles hanging over their head for the rest of their wiki-career. The intent is not to prevent action being taken against an editor who is messing up at high speed and high volume - we'd just have to do that the usual way, at ANI. -- asilvering (talk) 20:53, 23 January 2026 (UTC)\\[\\] </p>\n<p>I would support this, but the direct encouragement of LLM usage - Grok especially - still really worries me, and I do feel like @7804j inherently doesn't see the moral issue with using LLMs on Wikipedia. The outcome of Wikipedia:Village_pump_(policy)#Seeking_clarification_on_WP:NEWLLM_regarding_human-reviewed_translations will have a direct bearing on a OKA proposal going forward, since the LLM usage is integral to the OKA workflow. <strong>qcne</strong> (talk) 20:59, 23 January 2026 (UTC)\\[\\] </p>\n<p>Yes, and that would require a different kind of discussion, as you note. the moral issue with using LLMs on Wikipedia is not something we are going to reach consensus about in time to usefully deal with the issues raised here about unverifiable content being added to Wikipedia, possibly at scale, by a team of editors. And we can solve that one the way we've always solved it - with blocks - and authorize presumptive deletions to make that less of a burden. -- asilvering (talk) 21:10, 23 January 2026 (UTC)\\[\\]</p>\n<p>Nor does he need to. There is no consensus that there is a moral issue with using LLMs on WP Czarking0 (talk) 04:29, 11 February 2026 (UTC)\\[\\]</p>\n<p>I think this is a really useful proposal. I do think rule 1 might be slightly too harsh; this effectively blocks translation of any content referenced to offline sources. It seems to me that there are two qualities I want from OKA translator content: (a) translated content is cited to the same source as is used in the source article <em>or</em> a new source verified by the editor, (b) if cited to an offline source, the cited source \"seems reasonable\" (e.g. sufficient bibliographic metadata to find the source in principle, title seems related, etc.). But I think your proposal has the significant advantages of concision and clarity. Cheers, Suriname0 (talk) 21:23, 23 January 2026 (UTC)\\[\\] </p>\n<p>@Suriname0, it is my position that, since there are significant concerns about hallucinations and AI use with the OKA translators as a group, we've lost the ability to assume accuracy when it comes to offline sources. I wouldn't say it effectively blocks translation - someone with access to the offline source via a library, WP:LIB, WP:RX, or other methods could still translate those parts. And these days that does include quite a lot of sources that have previously been \"offline only\". -- asilvering (talk) 21:43, 23 January 2026 (UTC)\\[\\] </p>\n<p>Yes, true! I was counting WP:LIB + shadow libraries as functionally online these days, but especially with resource requests it's probably quite reasonable. It would probably be straightforward to generate a resource request for offline sources before translation begins on an article: \"I want to translate the French article fr:So-and-So, which refers to Some Encyclopédie, pages 54 &amp; 65, and I'd like to verify the content before I include it on English Wikipedia, thanks!\". Suriname0 (talk) 22:11, 23 January 2026 (UTC)\\[\\]</p>\n<p>I disagree that it's harsh. Every editor is responsible for complying with WP:V and should verify all content they add to the encyclopedia. voorts (talk/contributions) 23:26, 23 January 2026 (UTC)\\[\\]</p>\n<p>Yes, I agree except rule 1 needs tweaking. The point is to control for AI hallucinations, not to add extra burden on translators, one that even WP:AFC does not. Let me quote from WP:AFCSTANDARDS: \"Avoid declining an article because it correctly uses general references to support some or all of the material. The content and sourcing policies require inline citations for only four specific types of material, most commonly direct quotations and contentious material about living persons.\" It does not make sense to require translators in general, or even OKA in particular, to be required to meet higher standards than what we expect from an average new editor. In other words, we accept, through AfC, articles that are undeerreferenced (which I am not very happy about, but that's how the rules are written). Now, AfC rules don't seem to be updated to AI era, and we are indeed dealing with a different issue here - lazy translators not checking for AI hallucinations. So yes, OKA translators (and translators in general) need to check for them, and if they fail to do so, blocks are in order (just like for any editor who adds AI hallucinations to Wikiepdia). But no, requirement to check <em>all</em> sources, which effectively means checking even offline publications, is way above and beyond what is required from anyone outside FA and some GA level content. It would be an extremely unfair burden, and one that could be gamed easily by using AfC system, meeting their minimum standards, and just stripping article from most references - hardly an outcome we want. PS. Yes, I think we should tighten AFCSTANDARDS, but that's a different discussion. Piotr Konieczny aka Prokonsul Piotrus| reply here 04:02, 24 January 2026 (UTC)\\[\\] </p>\n<p>But no, requirement to check all sources, which effectively means checking even offline publications, is way above and beyond what is required from anyone outside FA and some GA level content. Would a compromise such as that proposed by Suriname0 above be acceptable? My worry is, for example, if an AI model adds spurious citations to offline sources (which it couldn't have accessed). Requiring to check sources for citations that you add (that weren't there in the original) should presumably be a baseline. Chaotic Enby (talk · contribs) 04:30, 24 January 2026 (UTC)\\[\\] </p>\n<p>This is all a good best practice, but it is not something we can reasonably require, and again, it would be more than we require from AfCs (or from a regular translator). What we need is a requirement to check for AI hallucinations - that I fully support. But any requirement that a translator should verify content from another Wikipedia is beyond what we have ever required, and it would make translating much more burdensome than writing an article from scratch. Piotr Konieczny aka Prokonsul Piotrus| reply here 04:36, 24 January 2026 (UTC)\\[\\] </p>\n<p>Checking sources that you add yourself (whether directly or through an AI model) that weren't in the original text is absolutely something we can reasonably require. When you publish an edit, you take responsibility for the content you add. Chaotic Enby (talk · contribs) 04:42, 24 January 2026 (UTC)\\[\\] </p>\n<p>@Chaotic Enby I agree with that, as long as \"sources you add yourself\" does not involve copying within Wikipedia (including cross project). We do have WP:AGF and when few weeks ago I for example split the List of lost literary works from the main lost literary work I can assure you I did not check a single source, nor do I feel I should've had; nor did I check the refs in the List of Fading Suns books I likewise split a few days ago. Such splits, just like translations, are, IMHO, mechanical actions, not proper content creation that requires the creator to take responsibility for what's in the text body (as long as they do their work correctly - and being lazy and adding AI hallucinations is certainly not something we should condone, that's below acceptable levels of quality, as it makes new content worse, whereas proper translation or split should produce content that's identical to what already has been accepted by the community, and thus falls within AGF). Piotr Konieczny aka Prokonsul Piotrus| reply here 14:27, 24 January 2026 (UTC)\\[\\]</p>\n<p>@Piotrus, we're talking about editors who have repeatedly come to the attention of patrollers for creating articles that contain hallucinations. It is time to add \"extra burden\" to these translators. We are not talking about AFC broadly or translation broadly, but this specific subset of editors. -- asilvering (talk) 04:39, 24 January 2026 (UTC)\\[\\] </p>\n<p>@Asilvering Fair. What we need is a list of problematic editors, an understanding of the percentage they make of OKA editors, how that percentage compares to equivalent projects in good and bad standing (i.e. what percentage is a red line?), and then we can figure out what extra burden to add to them (and to be clear, I fully support a requirement for any and all translators - not just OKA ones - to check for and remove any AI hallucination, and to block repeated offenders). Piotr Konieczny aka Prokonsul Piotrus| reply here 14:58, 24 January 2026 (UTC)\\[\\]</p>\n<p>so we end up with \"OKA screwed up again\" not \"editor x screwed up again\".</p>\n<p>As it should be. This is a structural and process problem, not a problem of a few bad actors. It's hard to blame Editor X for using AI when they were specifically told to use AI. Gnomingstuff (talk) 21:30, 23 January 2026 (UTC)\\[\\]</p>\n<p>This seems promising. Is there, though, a specific reason why you chose a four-strike rule as opposed to three strikes? Also, it would be useful to have a table or something along those lines of OKA editors and the articles that they make so that tracking/auditing is made easier, as per GnomingStuff. Then there’s also the LLM translation RfC, which covers OKA. If the RfC passes, would this proposal be made redundant or would the measures here be taken as precedent? <strong>I</strong>seulttalk to me 21:35, 23 January 2026 (UTC)\\[\\] </p>\n<p>This sheet and Category:Articles translated by an OKA editor should be helpful for that purpose. Chaotic Enby (talk · contribs) 21:37, 23 January 2026 (UTC)\\[\\]</p>\n<p>@Iseult, I went with four strikes since our normal set of escalating warning templates have four levels. -- asilvering (talk) 21:38, 23 January 2026 (UTC)\\[\\]</p>\n<p>While I strongly support this as a baseline from which we can start, and I believe the question of \"moral issues\" of LLMs is ultimately a red herring, there are a few more things I would like to see addressed. ~~For example, the original remedies suggested transparency in the OKA process: the targeted donations program is an aspect of this, and our paid-editing policy would require these to be disclosed.~~ Additionally, large-scale editing campaigns (such as lead format changes in many articles at once) should require explicit on-wiki consensus rather than OKA-specific consensus. Chaotic Enby (talk · contribs) 21:36, 23 January 2026 (UTC)\\[\\] </p>\n<p>+1 Ethmostigmus 🌿 (talk | contribs) 01:19, 24 January 2026 (UTC)\\[\\]</p>\n<p>Struck the part about targeted donations which has been clarified. I'm already much more reassured about the situation. Chaotic Enby (talk · contribs) 07:48, 24 January 2026 (UTC)\\[\\]</p>\n<p><strong>Support</strong>. This is a narrowly tailored solution to the problem. Technically #2 should be a 5-strike rule since strike 4 is the final warning. voorts (talk/contributions) 23:31, 23 January 2026 (UTC)\\[\\] </p>\n<p>Thanks @Voorts, point taken. I'll clarify that. -- asilvering (talk) 04:41, 24 January 2026 (UTC)\\[\\]</p>\n<p><strong>Weak support</strong> as bare-minimum. This is a WP:CIR issue: every editor, regardless of what motivates them to edit and what tools they use, is responsible for verifying the information they put on Wikipedia. If an editor is utilising a tool that has a higher likelihood of making mistakes (hallucinating references or substantially changing the meaning of translated text), they should be extra vigilant in verifying both references and the integrity of the text. This is the absolute minimum level of competency we should expect of <em>all</em> editors, volunteer or paid.<br>\nRealistically, this is not an issue that can be fully addressed with onwiki sanctions alone - what's really needed is improvements to internal oversight and procedure at OKA, which is obviously out of our hands here, but I hope this thread has given 7804j food for thought about where improvements can be made. Currently, issues identified with OKA editors' work internally and onwiki seem to exist in separate streams: issues identified internally are, presumably, dealt with internally, and issues identified onwiki may be addressed onwiki <em>or</em> internally. I think we need better transparency here, so both OKA editors and the wider community can be on the same page when issues arise to allow both groups to better identify ongoing issues. Ethmostigmus 🌿 (talk | contribs) 01:16, 24 January 2026 (UTC)\\[\\] </p>\n<p>I think we also need further clarification, transparency, and community consensus on the paid editing aspect of OKA. Chaotic Enby's suggestions above regarding appropriate disclosure of targeted editing and onwiki consensus for large-scale changes are a good start, but I have a few other things I think are worth discussing further.</p>\n<ul>\n<li>OKA's translation taskforce page states Since OKA translators are paid editors, according to WP:FCOI, all their contributions to English Wikipedia are considered to have a conflict of interest. As a result, they are not allowed to create new articles here directly, but must use the articles for creation (AfC) process - this is good, but the targeted donations add an extra wrinkle - I think a distinction should be made between edits directly related to and not directly related to these donations when declaring COI. As a starting point, if not being done already (I don't <em>think</em> it is, but apologies if I am incorrect!), I think specific disclosures (\"I was paid to edit this article on <div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>t</mi><mi>o</mi><mi>p</mi><mi>i</mi><mi>c</mi><mi>A</mi></mrow></math></div> as part of my paid editing work funded by a donation from <div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>d</mi><mi>o</mi><mi>n</mi><mi>o</mi><mi>r</mi><mi>a</mi><mi>s</mi><mi>s</mi><mi>o</mi><mi>c</mi><mi>i</mi><mi>a</mi><mi>t</mi><mi>e</mi><mi>d</mi><mi>w</mi><mi>i</mi><mi>t</mi><mi>h</mi><mi>t</mi><mi>o</mi><mi>p</mi><mi>i</mi><mi>c</mi><mi>A</mi></mrow></math></div>, see <div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>i</mi><mi>n</mi><mi>s</mi><mi>e</mi><mi>r</mi><mi>t</mi><mi>l</mi><mi>i</mi><mi>n</mi><mi>k</mi><mi>t</mi><mi>o</mi><mi>O</mi><mi>K</mi><mi>A</mi><mi>t</mi><mi>a</mi><mi>r</mi><mi>g</mi><mi>e</mi><mi>t</mi><mi>e</mi><mi>d</mi><mi>d</mi><mi>o</mi><mi>n</mi><mi>a</mi><mi>t</mi><mi>i</mi><mi>o</mi><mi>n</mi><mi>s</mi><mi>p</mi><mi>a</mi><mi>g</mi><mi>e</mi></mrow></math></div>\") would be much better than the standard COI disclosure. A list of edits made at the direction of donors would also be helpful, if it doesn't exist already.</li>\n<li>OKA's Meta page states that Individuals who donate over CHF 5,000 per year are eligible to join OKA as formal members with voting right in the annual assembly: @User:7804j, can you give us a bit of background on what this entails? What influence do donors who become formal members have, and what kind of matters would they vote on at these assemblies? Is there a public list of these members and their affiliations (ie. primarily editor/translator vs primarily donor)? This system appears quite opaque to those of us outside OKA, which naturally raises suspicion given that this involves paid editing, so I think transparency is in all of our best interests.</li>\n</ul>\n<p>Ethmostigmus 🌿 (talk | contribs) 02:44, 24 January 2026 (UTC)\\[\\] </p>\n<p>We so far never received any targeted donation, which is why we haven't disclosed any yet. This statement was more to keep the door open for any such situation in the future. If we ever do, we would make the list of such donations, their scope and requirements, public and clear. However, I would be against requiring it in the edit summary, primarily because it makes it extremely complex operationally for us to manage. Also in practice, I expect (and will require) that such targeted donations be sufficiently broad to avoid undue influence. For example, a donation could be for \"any article related to Switzerland\", so I wouldn't want people to think that because a translator worked on an article for a Swiss company, that this is because that particular company paid for it, when in fact it was just because they were asked to work on articles about Switzerland.</p>\n<p>Regarding the donation of 5,000 CHF and above: this is linked to our legal status as an association, which requires a mechanism for adhesion. We would vet such members, so not everyone who donates 5000 will automatically be able to join as a member. We can disclose these members if the community is interested, but so far they only consist of myself and people I am close with who donated such amounts (none of them are active on Wikipedia). 7804j (talk) 07:50, 24 January 2026 (UTC)\\[\\]</p>\n<p><strong>Moral support</strong>. This is a sensible proposal, grounded in existing English Wikipedia policy. I'm not certain it is necessary, nor am I certain it will fix what it aims to fix, but it's better than the pitchfork mob approach above.</p>\n<p><strong>Necessary?</strong> In my experience, 7804j has been very receptive to specific, actionable feedback, including rapidly firing an OKA editor who was falsifying sources (link to discussion). In that case, it took only one warning for action from OKA itself, far fewer than the four warnings proposed here. It is alarming that it seems nobody went to 7804j's talk page to raise concerns about the LLM rewrites of existing articles before bringing the issue to this noticeboard, which would have been a far more efficient way of resolving this.</p>\n<p><strong>Sufficient?</strong> The challenge with detecting issues in OKA drafts is that they tend to be very long and largely unproblematic. In the case I linked above, @MCE89 detected serious issues, but those same issues had been missed by several other reviewers – it's not in the purview of an AfC/NPP reviewer to meticulously check source–text integrity, esp. for OKA drafts which are usually fine (source: my own AfC reviews of OKA content, which often include spot-checks, and also CE's analysis below), when there is so much actual crap that needs their attention. This proposal theoretically makes it easier to block OKA editors, but will we actually do that? Will we actually check any more than we currently do? I doubt it. <strong>Toadspike </strong></p><div class=\"math-box\"><strong><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>T</mi><mi>a</mi><mi>l</mi><mi>k</mi></mrow></math></strong></div> 02:58, 24 January 2026 (UTC)\\[\\]<p></p>\n<p><strong>Weakest possible support</strong> per my concerns here, I do not want OKA workers punished as a consequence of poor management. <strong>ClaudineChionh</strong> (<em>she/her</em> · talk · email · global) 04:40, 24 January 2026 (UTC)\\[\\]</p>\n<p><strong>Mostly supportive.</strong> As others mentioned, this proposal imposes stricter criteria on OKA editors than on regular editors, but I can nonetheless support it as the requirements and process listed is very sensible, and already consistent with our own guidelines.</p>\n<p>The only part I disagree with, however, is to require that the content of every offline source is checked. As others have pointed out, it is often very difficult or impossible to check the content of a book, especially when these aren't digitized or extremely expensive. I agree that translators should be required to check the content of any <strong>new</strong> offline source they add to ensure it isn't hallucinated, and of any online source, but in the context of a translation of offline references, I think it should be sufficient to ensure that the reference was correctly placed and translated and thus not hallucinated. As others mentioned, this is the same standard that we would hold non-OKA translators to.</p>\n<p>Without this exception, I expect that we would need to throw out so many articles or paragraphs that translation would become effectively impossible 7804j (talk) 07:29, 24 January 2026 (UTC)\\[\\]</p>\n<p><strong>Moral oppose</strong> - I'll just say it: I don't think we should be encouraging people, let alone <em>paying</em> people, to add AI content to Wikipedia at all. I'm not sure why that's being framed as an extreme stance; I suspect if we asked readers whether they think people should be paid to add AI content to Wikipedia, the median response would be closer to \"no, what the fuck?\". Scapegoating a handful of editors for doing exactly what they were told to do is closer to a \"pitchfork mob\" stance, imo.</p>\n<p>This isn't analogous to WikiEdu because WikiEdu doesn't tell students to use AI -- they discourage it in their training guide, and they've been receptive to feedback on the specifics. Obviously some students are going to do it anyway, but that's entirely on them, it's not part of their instructions. Gnomingstuff (talk) 09:26, 24 January 2026 (UTC)\\[\\] </p>\n<p>I suspect if we asked readers whether they think people should be paid to add AI content to Wikipedia, the median response would be closer to \"no, what the fuck?\" Are you referring to the same readers (i.e., consumers of information online) who are slowly shifting to using LLMs in lieu of search engines and coming to trust LLMs more and more? This opposition to LLM use within the community does not map onto the real world. voorts (talk/contributions) 21:42, 24 January 2026 (UTC)\\[\\] </p>\n<p>If it didn't map onto the real world, then Wikimedia wouldn't be running an enormous fundraising campaign about how Wikipedia is \"human-created\". Generally speaking, people don't do ad campaigns about unpopular opinions. Gnomingstuff (talk) 23:29, 24 January 2026 (UTC)\\[\\] </p>\n<p>Yeah its like they play both sides</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mn>4</mn></mrow></math></div> and it doesn't matter what advertising they do. Voorts is right here. \\~2025-38536-45 (talk) 17:06, 27 January 2026 (UTC)\\[\\] <p></p>\n<p>I think there are plenty of opinions that would entail both using LLMs as a source of information and also strongly opposing it for Wikipedia editing. Including the fact that it's one of the main <em>inputs</em> for LLMs, and you might not want such a thing to be filled with LLM <em>output</em>. Sesquilinear (talk) 03:24, 9 February 2026 (UTC)\\[\\]</p>\n<p>I don't think we should be encouraging people, let alone <em>paying</em> people, to add AI content to Wikipedia at all. This. This. A thousand times this. Stepwise Continuous Dysfunction (talk) 19:56, 25 January 2026 (UTC)\\[\\]</p>\n<p><strong>Support</strong> I will support essentially any plausibly actionable accountability mechanism for machine-assisted translations. This will add accountability from day 1. It will also allow for enforcement without having to file a complex LLM/machine translation case at ANI, which is sufficiently unpleasant that it deters reporting. NicheSports (talk) 15:28, 24 January 2026 (UTC)\\[\\]</p>\n<p><strong>Support</strong> - If they can't verify a source due to not having the original book/document, perhaps they could tag it so someone else can check?</p>\n<p>Since AI is prone to embellishments, I'm rather wary of just leaving those alone without any human verification.</p>\n<p>Otherwise, my gut feeling is that it's reasonable to expect that an editor who is being paid (even a little) and has access to additional training &amp; support when compared to the standard volunteer editor, should be held to a higher standard - especially when relying heavily on a tool that's broadly untested (this project was/is part of the test), constantly evolving and is well known to produce errors when it's not used correctly.</p>\n<p>We may loosen the reins as time goes on, but I feel that the project needs extra guardrails until we have enough data and long-term experience of the project to make that determination.</p>\n<p>~~BTW I was a bit surprised to see that the three Senior editors/Managers have around 5k, 3k and 2k edits respectfully, which isn't egregiously low but isn't as high as I was expecting to see for that position. That's not to say they're not competent as some of them clearly are, I was just expecting more experience (perhaps incorrectly).~~ Blue Sonnet (talk) 15:54, 24 January 2026 (UTC)\\[\\] </p>\n<p>To clarify regarding the senior editors: it's not based solely on the quantity of edits, but to a larger degree on factors such as the quality of their contributions, their ability to coach others and to improve our existing processes. Typically, once they receive that role, their edit counts tend to decrease as they spend more time on coaching. Also, note that OKA edits include more bytes on average than the typical edit, since they are often focused on very large articles. 7804j (talk) 16:08, 24 January 2026 (UTC)\\[\\] </p>\n<p>Noted, thank you! It wasn't a major issue so I made sure to put it in small font - it's good to know how that works. Blue Sonnet (talk) 16:17, 24 January 2026 (UTC)\\[\\]</p>\n<p>Are your workers compensated based on time, or based on how many words/bytes they produce? Apologies if you have already answered that elsewhere in this discussion. --Gurkubondinn (talk) 23:25, 24 January 2026 (UTC)\\[\\] </p>\n<p>7804j, I have not seen this addressed somewhere (but this discussion is very long so I may have missed it). Could you clarify this for me? --Gurkubondinn (talk) 12:21, 10 February 2026 (UTC)\\[\\] </p>\n<p>They are compensated solely based on the time they report, in order to encourage people to focus on quality over quantity. Though if a translator has an extremely very low output compared to others, we would discuss with them to understand what's going on (and also to catch fraud, as we've had instances of people reporting \"fake\" hours). 7804j (talk) 12:28, 10 February 2026 (UTC)\\[\\] </p>\n<p>7804j -- great, thank you! I was hoping that they were paid based on time (not based on word count), so that is good to hear. --Gurkubondinn (talk) 12:57, 10 February 2026 (UTC)\\[\\]</p>\n<p>@Blue-Sonnet, edit count for someone doing high-quality translations is going to be very low. Of all my edits, the ones that took the most time and effort - hours for each - were my translations. For someone exclusively focused on content work 5k edits is quite high indeed and represents a lot of work. -- asilvering (talk) 21:51, 24 January 2026 (UTC)\\[\\] </p>\n<p>Yeah, that makes a lot of sense - quality is definitely preferable to quantity. Blue Sonnet (talk) 23:59, 24 January 2026 (UTC)\\[\\]</p>\n<ul>\n<li>\n<p><strong>Oppose</strong> OKA should be barred from Wikipedia by any and all means possible until they forsake the use and even implicit encouragement of Grok, the flagship product of a deepfake porn site. Stepwise Continuous Dysfunction (talk) 19:48, 25 January 2026 (UTC)\\[\\]<br>\nThey did say above that they've stopped using Grok and switched to other LLM's. Blue Sonnet (talk) 01:32, 26 January 2026 (UTC)\\[\\]<br>\nNot according to their own documentation, which endorses it in multiple places. Stepwise Continuous Dysfunction (talk) 03:49, 26 January 2026 (UTC)\\[\\]<br>\nIt's not endorsed; in fact, the second of the three mentions clearly says that it's inferior to Claude and ChatGPT. The third mention might warrant rewording/clarification that Grok is not recommended. @7804j That said, there are plenty of other tools, and it's hard to keep recommendations of best tools up-to-date - I don't know if we even have any on wiki? (And saying 'don't use LLM' is increasingly pointless, although this all reminds me of the early days of Wikipedia, when teachers in schools were telling students don't use Wikipedia - how things have changed since, eh?).<br>\nAnyway, having read that page I think the general prompt could benefit from clear instructions about not adding anything that's not in the source text, and not omitting anything (being lazy) - the most common errors we see. Oh, and in the alread-present warning \"Ensure there are no AI hallucinations! In rare cases, LLMs may completely invent content that's not in the original text. If you introduce AI hallucinations, your Wikipedia risks being suspended\", isn't there a word missing after \"your Wikipedia\"? Account? Piotr Konieczny aka Prokonsul Piotrus| reply here 04:13, 26 January 2026 (UTC)\\[\\]<br>\nThanks for the suggestions! I have added them.<br>\n(by the way, feel free to directly edit our instructions if you feel like it -- as long as there are no controversial changes, I don't mind having others making edits) 7804j (talk) 20:48, 26 January 2026 (UTC)\\[\\]<br>\nAnything other than a full-throated rejection is an endorsement. That's how instruction manuals work: point people at a tool, and they will use it. Stepwise Continuous Dysfunction (talk) 21:43, 26 January 2026 (UTC)\\[\\]<br>\nIn either case, there is no reason to disallow use of Grok as a translation tool; it seems to preform reasonably where from what anyone can see - unless you are familiar with any reliable data that says otherwise? I doubt it is going to insert Musk POV or such into most tasks (and if its hallucinations are Musk-flavord, that doesn't make them any worse or better other AI hallucinations, which we want zero of here, regardless of which LLM model produces them). Piotr Konieczny aka Prokonsul Piotrus| reply here 06:43, 27 January 2026 (UTC)\\[\\]<br>\nI think the objection is more specifically endorsing Grok. But if they're not doing that anymore, that's a step in the right direction. (As far as \"introducing Musk POV,\" there have been some, uh, not great and hasty changes to Grok's system prompt, so that's actually a decent possibility.) Gnomingstuff (talk) 15:40, 27 January 2026 (UTC)\\[\\]<br>\nThe very first mention of Grok in the linked documentation page is, Even for technical articles, LLMs such as Claude, ChatGPT and Grok correctly translate &gt;90% of text and handles most formatting. That's not just permission to use Grok; it's actively saying that Grok is a good thing. The third of the three is a plain instruction: Go to https://claude.ai/ or chatgpt (or https://grok.com/). In between the two is Grok used our preferred tool, and some editors prefer it because it tends to respond better to instructions and has a higher daily cap; however, it's writing style is poorer, so it's not recommended. A mild disinclination to recommend because of writing style and not the fact of its being a propaganda machine from a revenge-porn company is... insufficient. Wikipedia is willing to deprecate sources; organizations that seek to contribute to Wikipedia should be willing to deprecate tools. Stepwise Continuous Dysfunction (talk) 23:20, 27 January 2026 (UTC)\\[\\]<br>\nAnd we still need more than IDONTLIKEMUSK to declare Grok bad for translation. In fact, I trust OKA (which provides a lot of data) more than POVed views here. Which LLMs to use should be based on data about their efficiency/error-rate or such, not on irrelevant anecdotes about how a given AI is used to create porn or whatever. Piotr Konieczny aka Prokonsul Piotrus| reply here 04:40, 28 January 2026 (UTC)\\[\\]<br>\nI don’t see why your first sentence is true: many people obviously would be happy with such a straightforward declaration. Maybe you mean that you personally object to it? But that’s obvious, you’ve been saying so at great length. \\~2025-31850-11 (talk) 11:12, 28 January 2026 (UTC)\\[\\]<br>\nI'm not arguing it's bad for translation, I'm arguing that it's bad for optics. Gnomingstuff (talk) 00:50, 31 January 2026 (UTC)\\[\\]<br>\nI am afraid I don't understand either what you or the IP mean. Piotr Konieczny aka Prokonsul Piotrus| reply here 02:30, 31 January 2026 (UTC)\\[\\]<br>\nIt has a non-trivial chance of being a PR nightmare for Wikipedia. Gnomingstuff (talk) 21:22, 6 February 2026 (UTC)\\[\\]<br>\nIf you mean, our treatment of volunteers who want to help, and how we criticize them for doing so, and demand impossible standards, then I can agree. Piotr Konieczny aka Prokonsul Piotrus| reply here 01:14, 12 February 2026 (UTC)\\[\\]<br>\nAre you deliberately trying to be obtuse? Gnomingstuff (talk) 04:15, 13 February 2026 (UTC)\\[\\]<br>\nYes, of course -- but really what is the point of this side-discussion anyhow, the establishment or not of consensus certainly doesn't depend on one of you convincing the other one. \\~2026-92659-0 (talk) 21:22, 13 February 2026 (UTC)\\[\\]</p>\n</li>\n<li>\n<p><strong>Support if necessary</strong> A tool is only as good as the user. Hold the users </p>\n</li>\n</ul>\n<p>Context: This section continues the discussion on OKA's practices, focusing on LLM usage, source verification, English proficiency, transparency, and specific comments from various editors. It also includes a debate about the use of Grok and its associated controversies.</p>\n<hr>\n<p>accountable for their edits. I see no problem with openly disclosed paid editing with reasonable actors. I too question whether it is necessary. Buffs (talk) 04:44, 4 February 2026 (UTC)\\[\\]</p>\n<ul>\n<li><strong>Moral oppose</strong> - I am perturbed by the attitude taken by OKA representatives and their defenders throughout this discussion, which seems to be an entirely backwards way to go about getting an initiative like this off the ground and essentially fait accompli, followed by wikilawyering. Frankly, if I'd seen the back-and-forth from January 20 occur as it was happening, I probably would have blocked 7804j on the spot for a combination of CIR, NOTHERE, and generally disrespecting our community's time, effort and ethos. I am left without confidence that they will do the right thing going forward. I also share concerns about exploiting international income differentials for the project, and the community dynamics of then expecting volunteer editors to police the exploited workers and effectively decide whether or not they get to continue their jobs. That having been said, I don't want to get in the way of potentially effective regulation if we can't get consensus for asking OKA to go back to the drawing board and start with a proposal to the community and/or WMF. signed, <em><strong>Rosguill</strong></em> 22:34, 4 February 2026 (UTC)\\[\\]</li>\n<li><strong>Moral oppose</strong>; I do not think these measures are sufficient, but something is better than nothing. I furthermore note that even community oversight of these edits will functionally be close to none, given the scale of editing. Every editor is responsible for the content of their edits. If they use tools known to misrepresent sources even outside of translation or to insert fabulist sources, the appropriate level of scrutiny is more, not less. At minimum I would like a three-strikes-and-out system wherein each instance (not just diff) of misrepresentation is a strike. <strong>I</strong>seulttalk to me 23:48, 4 February 2026 (UTC)\\[\\]</li>\n<li><strong>Support 2-4, oppose 1</strong>. I don't think I voted yet (and ummm it seems we are voting now? Shrug). My oppose of 1 is due to ambigious wording as discussed above, and I might support it if it's clear than that the requirement to check related to checking for a) added AI errors/halluciantions and b) ommissions (AI being lazy and not translating stuff), although b) is much less a problem. But the current wording is unclear enough it can be understood as a requirement for the translator to verify whether sources in the original article (even offline or likewise hard to access) support the original article, and that's excessive burden. In addition, I'd also add a prohibition to translate content that's LLM-generated, and a requirement to check if an article is LLM generated (at minimum, by looking at tags, but perhaps also by doing something else, like using ZeroGPT or looking at edit history? This would be a productive venue to discuss that I haven't really seen taken up by anyone here or elsewhere). --Piotr Konieczny aka Prokonsul Piotrus| reply here 01:39, 5 February 2026 (UTC)\\[\\]</li>\n<li><strong>Support</strong> Probably not enough, but a decent start. We need to do something, after some years of just worrying. Johnbod (talk) 22:56, 5 February 2026 (UTC)\\[\\]</li>\n<li>Support <strong>1</strong> as a general principle for all translations, not just OKA. Less sure about 2, individual admins can have their own discretion. Support <strong>3</strong>, somewhat addresses the reviewer time imbalance. No problem with 4, saves wikilawyering. CMD (talk) 05:06, 7 February 2026 (UTC)\\[\\]</li>\n<li>\n<p><strong>QUESTION</strong>: Is there no way for Wikipedia to obtain translations of texts, other than by reliance on LLMs (Large Language Models), with their shortcomings that have been discussed above? I am human, and over many years the accuracy of my translations for Wikipedia has practically never been questioned. There must surely be other bilingual human translators who are already satisfactorily participating in the Wikipedia project or who could be recruited to it. Nihil novi (talk) 07:02, 8 February 2026 (UTC)\\[\\]<br>\n@Nihil novi It's worth noting, first, that OKA was not using LLM until recently (when? good question), but shifted to using it - which I find unsurprising, as everyone and their dog is using LLMs for translations these days. I am not sure if going back to pre-LLM days is feasible - or doable. To me, asking someone to stop using LLMs tools in translation (responsibly, of course) is akin to asking them to stop using a word processor and go back to typewriter (or hand writing). Piotr Konieczny aka Prokonsul Piotrus| reply here 04:30, 9 February 2026 (UTC)\\[\\]<br>\nAgreed, an LLM can be a useful tool. Ultimately, a sentient human must be responsible for a translation. The question always comes back to the qualifications of the bilingual translator. Nihil novi (talk) 09:29, 9 February 2026 (UTC)\\[\\]<br>\n@Nihil novi That it does, no disagreement here. What puzzles me is that some folks seem to think the translator should also verify whether content in the original article is correct (supported by references cited there), which IMHO goes way beyond what one can reasonably expect the translator to do. Piotr Konieczny aka Prokonsul Piotrus| reply here 13:00, 9 February 2026 (UTC)\\[\\]<br>\nI agree. The translator's job is to translate a text. Investigation of the reliability of the text that is being translated is a responsibility of the article's authors. It is not a house painter's responsibility to investigate the structural stability of the house. Nihil novi (talk) 19:19, 9 February 2026 (UTC)\\[\\]<br>\n@Nihil novi, I disagree quite strongly on this point, since I believe it is each contributor's responsibility to do basic due diligence so that they aren't spreading misinformation. That doesn't mean that all translators should scrupulously verify each and every claim personally, but they should at a bare minimum have done enough cross-checking to make sure that what they're publishing is not substantially false. -- asilvering (talk) 04:32, 11 February 2026 (UTC)\\[\\]</p>\n</li>\n<li>\n<p><strong>OPPOSE.</strong> In general, the reliability of a text that a translator is rendering into the target language should already have been vetted prior to the text being translated. The only case when the translator can be held responsible for vetting the source-language text's reliability is if the text has been introduced into the article by the translator himself. Nihil novi (talk) 05:52, 11 February 2026 (UTC)\\[\\]</p>\n</li>\n</ul>\n<h3>Spot checks</h3>\n<p>\n</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>e</mi><mi>d</mi><mi>i</mi><mi>t</mi></mrow></math></div>\n<p></p>\n<p>I've decided to spot-check the first few articles in Category:Articles translated by an OKA editor translated from French (as that is my native language and I would be able to verify them more easily).</p>\n<ul>\n<li>16 May 1877 crisis: In #Initial hesitations between republic and monarchy, some offline sources were swapped without a given explanation, such as Duclert 2021, p. 122-132. being replaced with Houte 2014, p. 27. (another source that was being used somewhere else). Since both sources are offline and would be hard to access, the change being made without explanation is surprising, and I'm not especially confident about text-source integrity here. Besides this one small issue, the rest of the article is a very good translation, adapting the French Wikipedia article to the English Wikipedia's writing style while staying accurate to the original (which was a featured article).</li>\n<li>20,000 Colombian peso note: Very accurate, couldn't find any concerning divergences between the two versions (original in French was a good article).</li>\n<li>1824 Saint-Louis coal mine disaster: A few word swaps here and there, but again, great translation all things considered.</li>\n<li>1879 French Senate election: Now this is a very different case. While the entire article is marked as translated, and was added in a single edit, only half of it actually comes from the French original. The first two sections are fully new, with several new sources being cited, the main two being <em>Providential Men: History of a French Fascination</em> and <em>Dictionary of Ministers (1789–1989)</em>. The latter is cited in pages 45–46 for the one/two-round electoral system. The relevant pages talk about completely unrelated personalities (Étienne Clavière, Charles Cochon de Lapparent, Jean-Baptiste Collin de Sussy and Emmanuel Crétet), all dead long before the relevant senate election, or even the Third Republic itself! I couldn't find an online version of the former, but it appears to be a series of biographies of \"great men\", which doesn't seem especially consistent with it being used to source the organization of a senate election.</li>\n<li>1914 French mobilization: Despite the source being again a featured article, there are unsourced paragraphs, which were ported (and expanded!) in the translated version. Unsourced sentences like A small cadre of volunteer career officers and NCOs would be supplemented by millions of reservists, who had received training through their mandatory two-year-service at age 20 and periodic refresher trainings. appear from virtually nowhere (and is contradicted a few paragraphs below, where it is stated that the duration of military service varied as different laws were passed). Besides that, a few minor issues (the territorial army and its reserve being confused, <em>état de siège</em> being half-translated as state de siège, broken piped links like \"including Algeria\"), but nothing too dramatic.</li>\n</ul>\n<p>Conclusion? The average OKA translation is certainly good, but errors are still present, especially when the source article has unsourced paragraphs or empty sections, with text-source integrity being one of the main worries. Chaotic Enby (talk · contribs) 22:47, 23 January 2026 (UTC)\\[\\] </p>\n<p>@Chaotic Enby It's good to have some data. What you missed is identifying the specific translators responsible for the work. The one problematic article you found, 1879 one, is work of @Leeanah, and per WP:DUCK, I am concluding this particular translator (whose substandard work is what started this entire issue AFAIK) has been using LLM without checking for AI hallucinations. I would recommend pulling all her content from mainspace or at least tagging them with LLM warnings, and I would also ban that editor from any new submissions until they fix their past messes. Now, all the other articles were penned by different translators - which is a good sign, as it indicates most OKA translators are proofreading their work more carefully. So, really, all the evidence we have points to a single \"bad egg\" (although, of course, this is still just a sample, and it's possible more problems could be found - laziness is hardly a rare vice). Piotr Konieczny aka Prokonsul Piotrus| reply here 04:11, 24 January 2026 (UTC)\\[\\] </p>\n<p>This is just not accurate - did you click through and read any of the discussions qcne linked in his original post? The linked discussions involve multiple editors, as well as 7804j themself being forthright about the fact that a small number of OKA editors were not doing the right thing in certain cases. There <em>are</em> some problems with a small portion of OKA editors, and denying that problems exist does not fix them. Ethmostigmus 🌿 (talk | contribs) 04:33, 24 January 2026 (UTC)\\[\\] </p>\n<p>Someone should make a table, and clearly indicate what percentage of OKA editors have been found to be fall significantly below our standards, what action was taken by OKA and the community, and how that percentage compares to other projects such s Wiki Edu and so on. Otherwise we are talking apples, oranges, ILIKEITs and IDONLIKEITs. Piotr Konieczny aka Prokonsul Piotrus| reply here 14:18, 24 January 2026 (UTC)\\[\\]</p>\n<p>I agree that it seems like two editors here (Leeanah and Luizadnts) seem to have been the cause of quite a few of the issues raised throughout the thread. They had already received a formal warning, and it seems the articles here predate that warning. I've raised these specific points to them, and told them to no longer create any new articles until they have reviewed in depth <strong>all</strong> of the articles they have previously created. I have also asked them to keep a detailed internal tracker of their progress. It has been communicated to them that their ability to continue receiving OKA grants will be contingent on them being able to correctly review and fix all of their previously published content. 7804j (talk) 15:25, 24 January 2026 (UTC)\\[\\]</p>\n<p>Checked the first short article in the W's, Walecznych Street (note I do not speak Polish). Noticeably AI. Promotional in tone, but besides the AI compulsion to turn \"has\" into \"features,\" the original seems to have been somewhat promotional too. Some content has been removed, but mostly unsourced and/or trivial stuff, so that's fine. However, there are some meaning changes:</p>\n<p>Google Translate mentions 11 Walecznych Street having a \"glass staircase\" whereas the translated article says \"glazed staircase,\" which are two different things (\"glazed\" implies ceramic, wood, etc). I don't know which is more accurate -- \"glass\" does seem suspect, admittedly.</p>\n<p>The translation changes \"designed by Helena and Szymon Syrkus\" to <em>built</em> by them, which seems likely to be factually wrong, especially since the same word is repeatedly translated as \"designed\" elsewhere.</p>\n<p>The translation claims that 12 Walecznych Street is \"similar to the Kiltynowicz villa,\" but the Polish article seems to only claim that the clinker slabs are, not the whole house.</p>\n<p>The translation changes \"window openings\" for 37 Walecznych Street (per Google Translate again) to \"rounded windows\" out of nowhere. These don't look rounded to me (assuming this is the right house -- another angle seems to show a \"37\" marker on the wood house and it's the only wood-looking one, so I'm guessing it is -- and there aren't different windows out back or something)</p>\n<p>Gnomingstuff (talk) 10:10, 24 January 2026 (UTC)\\[\\] </p>\n<p>@Gnomingstuff Since I am rather fluent in Polish: 1) glass &gt; glazed, it's a staircase with, hmmmm, that actually gave me a pause. Big windows? Walls made of mostly glass? It's not made from glass, obviously. Literary translation would be \"glassy\", which obviously is not correct, it's late at night and I am blanking on the right terminology. Anyway, glazed is an error. 2) Yes, designed. 3) The use of these slabs on the ground floor, to be exact (as in, it's the use that's called similar, not the slabs - nor the whole house). 4) 36A not 37 - which is a good reminder that human beings like you and me are perfectly capable of making errors too. Anyway, the original text was indeed wrong (\"It features rounded windows, a curved corner, and a garden-facing terrace\"). The correct translation would be : \"Its characteristic features include windows*, a curved corner, and a garden-facing terrace\". *-&gt; Polish text says, literary, \"window openings\", which is unclear in Polish and seems to me (again, late night, and I am not an architecture expert or fan) to be just a pointlessly long ways of saying window. Anyway, yes, that translation had some minor problems, but whether there are LLM errors or human errors, I am not sure, and in either case, they are minor errors, not something I'd reprimand someone over. Piotr Konieczny aka Prokonsul Piotrus| reply here 14:13, 24 January 2026 (UTC)\\[\\] </p>\n<p>@Nihil novi How would you translate a \"przeszklona klatka schodowa\"? Piotr Konieczny aka Prokonsul Piotrus| reply here 14:16, 24 January 2026 (UTC)\\[\\] </p>\n<p>I'm at a disadvantage, as i don't know what \"OKA\" is, and I don't have access to the context in which \"<em>przeszklona klatka schodowa</em>\" appears. Among perhaps other possibilities, it might be a \"glass-enclosed staircase\" or \"glassed-in staircase\". Is there a link to the Polish text? Nihil novi (talk) 01:41, 25 January 2026 (UTC)\\[\\] </p>\n<p>Another possibility: a \"staircase with a skylight at the top\". Nihil novi (talk) 02:01, 25 January 2026 (UTC)\\[\\] </p>\n<p>The article is Walecznych Street, specifically we are talking about a building at Walecznych 11 which features, according to text, a \"przeszklona klatka schodowa\". Unfortunately I have yet to locate a picture :( Piotr Konieczny aka Prokonsul Piotrus| reply here 07:05, 25 January 2026 (UTC)\\[\\] </p>\n<p>The house is visible on Google Street View (link blocked by Wikipedia). While it's not obvious what the stairs themselves are like, the centre of the building, as seen from the street, is all window from ground floor lintel to top floor ceiling. Assuming the stairs are inside those windows, I'd call it a \"glazed stairwell\". Daveosaurus (talk) 07:18, 31 January 2026 (UTC)\\[\\] </p>\n<p>Hah, then I stand corrected - seems glazed is correct English. Live and learn, and also, it goes to show that what some calls an error is not always one. Piotr Konieczny aka Prokonsul Piotrus| reply here 11:18, 31 January 2026 (UTC)\\[\\] </p>\n<p>FWIW I'm not sure I agree with daveosaurus. 'Glazed' typically implies a layer of some kind of lacquer applied onto another surface. I would not usually use it for something 'encased in glass' or 'surrounded by windows' or the like, nor would I use it for something made of glass; that would be 'glassy' or simply 'glass' Athanelar (talk) 11:57, 31 January 2026 (UTC)\\[\\] </p>\n<p>See Glazing (window), which perhaps should be linked from \"glazed\". Glaze itself is a DAB with no primary topic. Donald Albury 15:36, 31 January 2026 (UTC)\\[\\]</p>\n<p>@Nihil novi For the full context, see the top of this very thread, then the spot check of an article translated from Polish. <strong>ClaudineChionh</strong> (<em>she/her</em> · talk · email · global) 04:06, 25 January 2026 (UTC)\\[\\] </p>\n<p>Thank you. Nihil novi (talk) 08:02, 25 January 2026 (UTC)\\[\\]</p>\n<p>@Oliwiasocz Could you take a look at the above, and make changes where needed? As highlighted, it seems like several of the issues are from the source article, so I'd recommend fixing both the source and the translation 7804j (talk) 15:16, 24 January 2026 (UTC)\\[\\] </p>\n<p>Why did this thread devolve (in my opinion) to just pointing out a few editors mistakes and cleaning them up? Or is it just yhis subsection? \\~2026-51814-4 (talk) 15:51, 24 January 2026 (UTC)\\[\\] </p>\n<p>This is a spot check, to see what the proportion of errors is on a random sample of articles. Of course, we can't physically check every single translation, so checking a sample of 5 or so allows us to see how frequent/rare errors are. Chaotic Enby (talk · contribs) 17:59, 24 January 2026 (UTC)\\[\\] </p>\n<p>thank you. that spot check is a redlink though maybe the article you are thinking of is under a different name. \\~2026-54906-2 (talk) 16:17, 25 January 2026 (UTC)\\[\\] </p>\n<p>Yep, surprised we don't have an encyclopedia article on it. Wikipedia:Spot checking sources is one of our project pages on how we apply that practice. Chaotic Enby (talk · contribs) 21:17, 25 January 2026 (UTC)\\[\\]</p>\n<p>Hi, concerning the mentioned issues:</p>\n<ol>\n<li>I used \"glazed\" because in wiktionary it was described as \"(architecture, construction) Having glass in the windows\". Maybe using \"stairwell with windows\" would be better here? The word in the Polish article seems to be correct as well, since the Polish version of wiktionary also describes it as \"one that has a large amount of glass surfaces\".</li>\n<li>Yes, it was supposed to be \"...built between 1935 and 1937 and designed by...\".</li>\n<li>The original translation may have sounded confusing, so I think it could be reworded to \"...slightly recessed ground floor clad in clinker tiles (similarly to the Kiltynowicz villa at 26 Katowicka Street)...\"?</li>\n<li>Yes, I have made a mistake with using \"rounded windows\". This phrase could be translated as \"window openings\" in a more technical way, but leaving just \"windows\" is completely fine. I also changed the phrase in the Polish article.</li>\n</ol>\n<p>Oliwiasocz (talk) 00:28, 25 January 2026 (UTC)\\[\\]</p>\n<p>Thanks for the clarifications on the Polish wording (and on the address as well, my mistake there) -- my gut sense was also that the original article was maybe leaning a bit much into architectural trivia.</p>\n<p>Not sure, though, how pointing out factual inaccuracies in an article so they can be fixed and the article can be improved, though, is \"devolving\" -- it's literally the whole point of a wiki. Really not sure why you thought I was reprimanding anyone; if I wanted to reprimand someone I would just do that, there's a reason I didn't single out the editor. Gnomingstuff (talk) 23:44, 24 January 2026 (UTC)\\[\\] </p>\n<p>Improving articles is all good. And I'll note that Oliwiasocz, who has translated many articles form pl wiki over the years, is, IMHO, a good example of positive value OKA has brought to the project. She has translated, among others, several of my articles, some of which were GA on pl wiki, and, after minor c/e, were recognized as GA here too. Piotr Konieczny aka Prokonsul Piotrus| reply here 07:08, 25 January 2026 (UTC)\\[\\]</p>\n<p>This discussion is too concentrated on Grok, and as such is damning by association. There is no proof that Grok is the biggest source of deepfake pics, and their existence is irrelevant. The question is how does Grok compare to the Content translation tool (semi protected I think in en, but available elsewhere) and google babblefish. I suspect that It's better than both. But it's a very low bar ,, Wakelamp (talk) d</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>@</mi><mo>−</mo><mi>@</mi></mrow></math></div>b 21:54, 4 February 2026 (UTC)\\[\\]<p></p>\n<h3>People that could be contacted</h3>\n<p>\n</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>e</mi><mi>d</mi><mi>i</mi><mi>t</mi></mrow></math></div>\n<p></p>\n<p>Should any of the following be contacted</p>\n<ul>\n<li>WMF grants Ch who gave the grant to OKA</li>\n<li>WMF US donation Stewardahip who received the Zimmerman grant.. Open knowledge association also has a similar name to the international Open Knowledge Foundation (opendata.ch).which also has links to Wikimedia,.</li>\n<li>WMF google as Zimmerman seems to work for google (Potential COI as google matched giving also gives money to WMF, and this money may have come from Google Ch matched giving</li>\n<li>WMF legal to contact the Swiss non profit regulator for similar information to the IRS 990 form.. Wakelamp (talk) d<div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>@</mi><mo>−</mo><mi>@</mi></mrow></math></div>b 13:34, 8 February 2026 (UTC)\\[\\]</li>\n</ul>\n<p>@Wakelamp, I'm not sure what you're asking here. Contacted why, and about what? -- asilvering (talk) 01:48, 9 February 2026 (UTC)\\[\\]</p>\n<p>@asilvering - Apologies for lack of clarity. @User:Gnomingstuff mentioned PR concerns, but I don't think the above discussion has addressed potential WMF governance and oversight responsibilities enough. Two Wikimedia entities are involved granting and receiving similar amounts in different countries, it is unclear whether this is linked to the Google (a partner) donation matching system, and there may be legal restrictions involving a grant recipient paying money into a third country. WMF should be informed Wakelamp (talk) d</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>@</mi><mo>−</mo><mi>@</mi></mrow></math></div>b 11:50, 10 February 2026 (UTC)\\[\\] <p></p>\n<p>I assume that if they gave a grant to the organization they are aware of the legal side already, and at any rate I'm not sure what this would be intended to accomplish. Gnomingstuff (talk) 13:44, 10 February 2026 (UTC)\\[\\] </p>\n<p>Not sure whether they are aware.</p>\n<ul>\n<li>the donation and the grants are looked after by separate departments, and this case also in two different organisations and involve the CEO of OKA and OKA,</li>\n<li>the application form has no option does not mention donations with matching grants. This is called round tripping or circular funding, and can cause concerns with |donor matching schemes and reallocation of tied funds</li>\n<li>The CH grant application form states that \"Projects that replace volunteer action by directly funding someone to create content (e.g. editing articles, uploading photos)\" are not eligible for funding, but OKA were funded.</li>\n</ul>\n<p>Wakelamp (talk) d</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>@</mi><mo>−</mo><mi>@</mi></mrow></math></div>b 12:58, 12 February 2026 (UTC)\\[\\]<p></p>\n<p>1. Should the WMF policy be changed on paid edits to exclude entities like OKA?</p>\n<p>OKA are not breaking any WMF policies. Although a WMF posted in 2025 called |Should I pay for a Wikipedia article?stated \"The Wikimedia Foundation does not offer Wikipedia editing services of any kind in exchange for fees, although it does seek charitable donations to keep Wikipedia running.\". but WMF grants are effectively subsidising paid editinh</p>\n<p>2. Grant funding of paid editor amd paid project manager's who edit</p>\n<p>The CH grant application form states that \"Projects that replace volunteer action by directly funding someone to create content (e.g. editing articles, uploading photos)\" ot eligible for funding, but OKA were funded</p>\n<p>3. Do they wish to discuss with OKA about statements on their website and their process</p>\n<ul>\n<li>\"Hundreds of thousands of high-quality articles are not translated.\", the articles they are translating are not high quality in terms of article class\"</li>\n<li>High Quality - their instructions \" \"Don't always translate the whole article. You are not required to replicate a source article's entire structure.\" sp wikipedia:disrupt</li>\n<li>They are mis-leading in their statements \"Science, technology, engineering, and Finance are lacking compared to topics such as History, Geography, and Humanities. \"</li>\n<li>They wish to create articles in \"Content from non-anglophone countries that is underrepresented.', but from PT, DE, FR, and ES</li>\n</ul>\n<p>4. OKA statistics</p>\n<ul>\n<li>\"So far, we have created 2,200+ new articles on Wikipedia, which generate 8m views per year.\"</li>\n</ul>\n<p>Current figure is 4896 - they point to their annual report for the 8 million views</p>\n<ul>\n<li>They have 12,000 articles planned</li>\n</ul>\n<p>5. WMF statistics</p>\n<ul>\n<li>OAK statistics are , but outreach statistics are very different 13.5K Articles Created, 47.1K Articles Edited, 124K Total Edits</li>\n</ul>\n<p>53 Editors, 74.8M Words Added, 738K References Added, 2.34B Article Views. 2740 Commons Uploads</p>\n<ul>\n<li>There is a mismatch between the editors listed on meta and the editors on the outreach dashboard</li>\n<li>The statistics from paid editing are counting towards the totals for the outreach program</li>\n</ul>\n<p>6. Voting</p>\n<ul>\n<li>Some of these editors are eligible to vote.</li>\n</ul>\n<p>7. AI</p>\n<p>The WMF AI strategy document specifies that AI must be open weight. They are not using the mostly paid open weight models, and there is no record of which one was used Wakelamp (talk) d</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>@</mi><mo>−</mo><mi>@</mi></mrow></math></div>b 14:03, 12 February 2026 (UTC)\\[\\]<p></p>\n<h3>Move this discussion as a wikipedia space page</h3>\n<p>\n</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>e</mi><mi>d</mi><mi>i</mi><mi>t</mi></mrow></math></div>\n<p></p>\n<p>Long discussions become confusing quickly, as points are repeated or questioned. Could we create this as a page and then use our editing skills to create a summary, then get consensus on that? Wakelamp (talk) d</p><div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>@</mi><mo>−</mo><mi>@</mi></mrow></math></div>b 13:06, 12 February 2026 (UTC)\\[\\]<p></p>\n<ul>\n<li><strong>SUPPORT.</strong> Nihil novi (talk) 03:01, 13 February 2026 (UTC)\\[\\]</li>\n</ul>\n<p>Makes sense. Should be considered for being standard practice even for such long discussions \\~2026-10381-58 (talk) 19:25, 15 February 2026 (UTC)\\[\\] </p>\n<p>I am going to use a number of AIs - the irony - to create a starting point article for the community to look at and QA</p>\n<ul>\n<li>based on @Qcne original concerns, and suggested remedies, a few additional ones</li>\n<li>Then use AI to create a table showing references to short codes and points</li>\n<li>then use AI to create a collapsed fishbone diagram Wakelamp (talk) d<div class=\"math-box\"><math xmlns=\"http://www.w3.org/1998/Math/MathML\" display=\"block\"><mrow><mi>@</mi><mo>−</mo><mi>@</mi></mrow></math></div>b 13:20, 20 February 2026 (UTC)\\[\\]</li>\n</ul>\n<p>If you do this, please identify the exact models, model versions, and prompts you use to generate these. <strong>ClaudineChionh</strong> (<em>she/her</em> · talk · email · global) 03:53, 21 February 2026 (UTC)\\[\\] </p>\n<p>Use Grok. Popcorn's on me. Piotr Konieczny aka Prokonsul Piotrus| reply here 12:06, 21 February 2026 (UTC)\\[\\] </p>\n<p>Double dare accepted :-) Version 1. It's on meta because I am not certain what happens now if I mention grok on a sandbox page,</p>\n<p>Section B,C are giving some inaccurate result.</p>\n<p>Prompt</p>\n<p>Create the following</p>\n<p>Please check for AI hallucinations, and double check links .</p>\n<p>A..Headings : summary of discussion</p>\n<p>Create a collapsable sortable mediwiki markdown table suitable for posting to en Wikipedia format of the following discussions</p>\n<p>OKA discussion on the English Wikipedia Administrators' noticeboard thread titled approximately \"OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles\" (January–February 2026).</p>\n<p>This is not a replacement for editor discussion — Wikipedia's strength lies in </p>\n<p>Context: This section features further discussion on OKA's operations, including arguments about the fairness of stipends, the use of AI, the role of the founder, and proposals for stricter rules for OKA translators. It also touches upon the broader community's reaction and the potential impact of OKA's activities.</p>\n</details><details><summary>Thinking</summary>\n<p>The user wants me to identify every editor who contributed to this AN discussion and summarise their positions. Let me go through the thread carefully and compile this.</p>\n</details>\n<h2>Editors and their positions in the OKA AN discussion</h2>\n<p>Below is a table of every identifiable editor who contributed substantively to the thread, along with a summary of their position. Editors are listed roughly in the order they first appear.</p>\n<table>\n<thead>\n<tr>\n<th>Editor</th>\n<th>Position summary</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td><strong>Qcne</strong></td>\n<td>Thread opener. Documented the long history of problems with OKA translations and proposed remedies (incremental edits, on-wiki consensus for campaigns, transparency on QA). Grew increasingly critical of 7804j's evasive and LLM-drafted replies. Ultimately called for OKA editors to be <strong>prohibited from directly editing articles</strong> <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Johnbod</strong></td>\n<td>Long-standing critic of OKA quality. Described the problem as \"long-running and serious\", noted 7804j's intermittent engagement and poor QC. Sceptical that proposed remedies would work with the current set-up, but ultimately <strong>supported</strong> asilvering's proposal as \"a decent start\" <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>DoubleGrazing</strong></td>\n<td>Noted they had accepted some OKA drafts at AfC that seemed fine, but acknowledged they hadn't specifically looked for MT problems and questioned whether detecting them was within AfC's remit <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>WhatamIdoing</strong></td>\n<td>Pushed back on the assumption that translators can't work in multiple language pairs; noted multilingualism is normal in many parts of the world. Also contributed a lengthy (and controversial) aside on CSAM statistics <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Clovermoss</strong></td>\n<td>Briefly noted that language fluency is a spectrum and that limited fluency can restrict translation direction <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Cremastra</strong></td>\n<td>Agreed with the fluency-is-a-spectrum point. Later made a sardonic remark about AI proponents not doing cleanup work <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>CMD</strong></td>\n<td>Flagged the incident where 7804j detagged obviously LLM-generated text with fabricated sources. Supported asilvering's proposal point 1 (as a general principle for all translations) and point 3 <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>7804j</strong> (OKA founder)</td>\n<td>Defended OKA throughout. Argued errors are statistical outliers in a large project; described QA processes (warnings, grant termination); disclosed stipend amounts and organisational structure. Acknowledged using LLMs to draft his own discussion replies. Receptive to a peer-review pilot and some process changes, but pushed back on characterisations of exploitation and on requirements to verify offline sources. Dismissed the thread as unrepresentative of broader community sentiment <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>ExtantRotations</strong></td>\n<td>Highlighted that OKA's training guide instructs translators to create multiple email accounts to circumvent LLM usage caps. Questioned the \"grant not employment\" framing and quoted OKA's own website about maximising impact by recruiting in low-cost-of-living countries <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>ClaudineChionh</strong></td>\n<td>Asked detailed, structured questions about onboarding, selection, supervision, and training. Criticised 7804j for not seeking community approval before launching OKA. Expressed disgust at the gig-work arrangement and power dynamics, but gave <strong>weakest possible support</strong> to asilvering's proposal out of concern for OKA workers <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Iseult</strong></td>\n<td>Raised the general-to-native-language translation norm. Concerned about the scale mismatch (few volunteers policing many paid editors). Called for a <strong>three-strikes-and-out</strong> system and gave a <strong>moral oppose</strong> to asilvering's proposal as insufficient <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Gurkubondinn</strong></td>\n<td>Supported prohibiting OKA from direct article editing. Questioned compensation practices, asked 7804j to stop using LLM-drafted replies, and pressed for clarity on onboarding payment terms <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Chaotic Enby</strong></td>\n<td>Probed onboarding details and labour conditions (independent contractor status, low pay). Conducted <strong>spot checks</strong> of French-translated articles and found errors (hallucinated sources, unsourced expansions). Supported asilvering's proposal plus additional transparency measures (on-wiki consensus for mass campaigns). Later struck concern about targeted donations after 7804j clarified none had been received <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Nickoka</strong></td>\n<td>Mentioned by 7804j as co-manager of OKA (handles operations and finances). Not an active on-wiki editor; did not post in the thread <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Piotr Konieczny aka Prokonsul Piotrus</strong></td>\n<td>Most prolific defender of OKA alongside 7804j. Argued the project is a net positive; errors are marginal; accused critics of fearmongering and AGF violations. Opposed requiring verification of offline sources as excessive. <strong>Supported</strong> asilvering's proposal points 2–4 but <strong>opposed</strong> point 1 as ambiguously worded. Defended LLM use in drafting talk-page replies and pushed back strongly on the Grok backlash <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Gnomingstuff</strong></td>\n<td>Raised the PR/optics risk of using Grok specifically. Conducted a spot check of the Polish article <em>Walecznych Street</em> and found several meaning-changing errors. Gave a <strong>moral oppose</strong> to asilvering's proposal, arguing the fundamental problem is paying people to add AI content at all <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Ethmostigmus</strong></td>\n<td>Offered a nuanced view: acknowledged OKA's value but stressed the need for better internal oversight and transparency, especially around paid-editing disclosure and targeted donations. Gave <strong>weak support</strong> to asilvering's proposal as a bare minimum <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Hex</strong></td>\n<td>Criticised OKA's use of Grok, calling it a \"CSAM generator\" and questioning the judgement behind using it <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>LuniZunie</strong></td>\n<td>Strongly critical of 7804j's AI-generated replies and lack of respect for community wishes. Wanted more proactive, pre-emptive community engagement from OKA rather than after-the-fact discussions <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>asilvering</strong></td>\n<td>Author of the <strong>formal four-point proposal</strong> (verify all additions; four-warning block threshold within six months; presumptive deletion of blocked editors' unverified content; prospective application only). Defended the verification requirement against objections about offline sources <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Suriname0</strong></td>\n<td>Supportive of asilvering's proposal but suggested point 1 might be slightly too strict for offline sources; proposed a \"reasonable plausibility\" standard for citations that can't be directly checked <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>voorts</strong></td>\n<td><strong>Supported</strong> asilvering's proposal. Argued every editor is responsible for WP:V compliance; technically suggested it should be a five-strike rule. Also pushed back on anti-LLM sentiment as out of step with real-world trends <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Toadspike</strong></td>\n<td>Investigated OKA stipend levels against World Bank income data and concluded the exploitation charge was not well-founded. Gave <strong>moral support</strong> to asilvering's proposal but questioned whether it would actually change enforcement in practice <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Pi.1415926535</strong></td>\n<td>Firmly agreed OKA editors should be prohibited from direct article editing; criticised 7804j for launching OKA without community input and for not addressing major concerns <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Stepwise Continuous Dysfunction</strong></td>\n<td>Agreed with prohibiting direct editing. Later <strong>opposed</strong> asilvering's proposal, demanding OKA be barred entirely until it drops all use/endorsement of Grok <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Blue Sonnet</strong></td>\n<td>Asked practical questions about escalation contacts, ongoing QA, and post-probation oversight. <strong>Supported</strong> asilvering's proposal, emphasising that paid editors with extra training should be held to a <em>higher</em> standard <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>NicheSports</strong></td>\n<td><strong>Supported</strong> asilvering's proposal as providing day-one accountability and lowering the barrier to enforcement compared to a full ANI case <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Rosguill</strong></td>\n<td><strong>Moral oppose</strong> to asilvering's proposal: said they would have blocked 7804j on the spot for CIR/NOTHERE-type behaviour; concerned about exploiting income differentials and the dynamics of volunteers policing paid workers. Would prefer OKA to go back to the drawing board <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">accountable for their edits. I see no problem with openly disclosed paid editing with reasonable actors. I too question whether it is necessary. Buffs (talk) 04:44, 4 February 2026 (UTC)\\[\\] * **Moral oppose** \\- I am perturbed by the attitude taken by OKA representatives and their defenders throughout this discussion, which seems to be an entirely backwards way to go about getting an initiative like this off the ground and essentially fait accompli, followed by wikilawyering. Frankly, if ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Nihil novi</strong></td>\n<td>Asked why human translators aren't used instead of LLMs. Agreed translators shouldn't have to verify original-article sourcing. <strong>Opposed</strong> asilvering's proposal on the ground that source reliability is the original author's responsibility, not the translator's <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">accountable for their edits. I see no problem with openly disclosed paid editing with reasonable actors. I too question whether it is necessary. Buffs (talk) 04:44, 4 February 2026 (UTC)\\[\\] * **Moral oppose** \\- I am perturbed by the attitude taken by OKA representatives and their defenders throughout this discussion, which seems to be an entirely backwards way to go about getting an initiative like this off the ground and essentially fait accompli, followed by wikilawyering. Frankly, if ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Oliwiasocz</strong> (OKA translator)</td>\n<td>Responded to the spot-check critique of the <em>Walecznych Street</em> article, explaining and correcting the flagged issues <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Buffs</strong></td>\n<td><strong>Supported if necessary</strong>: saw no problem with openly disclosed paid editing by reasonable actors, but questioned whether the proposal was needed <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">accountable for their edits. I see no problem with openly disclosed paid editing with reasonable actors. I too question whether it is necessary. Buffs (talk) 04:44, 4 February 2026 (UTC)\\[\\] * **Moral oppose** \\- I am perturbed by the attitude taken by OKA representatives and their defenders throughout this discussion, which seems to be an entirely backwards way to go about getting an initiative like this off the ground and essentially fait accompli, followed by wikilawyering. Frankly, if ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Czarking0</strong></td>\n<td>Defended OKA against exploitation charges; argued the stipend criticism was paternalistic. An AI proponent who patrols NPP; accused critics of confirmation bias. Generally supportive of OKA continuing <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>2026-43687-3</strong> (IP/temp account)</td>\n<td>Called for OKA to be <strong>banned outright</strong>, characterising it as exploitative and neo-colonialist <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Wakelamp</strong></td>\n<td>Raised WMF governance and grant-oversight questions (circular funding, eligibility criteria, voting rights of paid editors). Proposed <strong>moving the discussion to a dedicated Wikipedia-space page</strong> and offered to create a structured summary using AI tools <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">accountable for their edits. I see no problem with openly disclosed paid editing with reasonable actors. I too question whether it is necessary. Buffs (talk) 04:44, 4 February 2026 (UTC)\\[\\] * **Moral oppose** \\- I am perturbed by the attitude taken by OKA representatives and their defenders throughout this discussion, which seems to be an entirely backwards way to go about getting an initiative like this off the ground and essentially fait accompli, followed by wikilawyering. Frankly, if ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Sesquilinear</strong></td>\n<td>Noted that people can consistently both use LLMs for information and oppose LLM-generated content on Wikipedia, since Wikipedia is a key <em>input</em> to LLMs <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Daveosaurus</strong></td>\n<td>Provided a Google Street View check of the <em>Walecznych Street</em> building, suggesting \"glazed stairwell\" was in fact correct <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">accountable for their edits. I see no problem with openly disclosed paid editing with reasonable actors. I too question whether it is necessary. Buffs (talk) 04:44, 4 February 2026 (UTC)\\[\\] * **Moral oppose** \\- I am perturbed by the attitude taken by OKA representatives and their defenders throughout this discussion, which seems to be an entirely backwards way to go about getting an initiative like this off the ground and essentially fait accompli, followed by wikilawyering. Frankly, if ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td><strong>Donald Albury</strong></td>\n<td>Linked to the <em>Glazing (window)</em> article to clarify terminology <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=accountable%20for%20their,to%20be%20%22&amp;text=built%20between%201935,and%20designed%20by&amp;text=%22.%203.%20The,reworded%20to%20%22&amp;text=slightly%20recessed%20ground,26%20Katowicka%20Street%29&amp;text=%22%3F%204.%20Yes%2C,strength%20lies%20in\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">accountable for their edits. I see no problem with openly disclosed paid editing with reasonable actors. I too question whether it is necessary. Buffs (talk) 04:44, 4 February 2026 (UTC)\\[\\] * **Moral oppose** \\- I am perturbed by the attitude taken by OKA representatives and their defenders throughout this discussion, which seems to be an entirely backwards way to go about getting an initiative like this off the ground and essentially fait accompli, followed by wikilawyering. Frankly, if ...</span></span></sup>.</td>\n</tr>\n<tr>\n<td>Various IPs/temp accounts (2026-47637-7, 2026-51814-4, 2025-38536-45, 2026-54906-2, 2025-31850-11, 2026-92659-0, 2026-10381-58)</td>\n<td>Brief interjections: some called for hatting tangential sub-threads, some questioned the direction of the discussion, one supported moving it to a dedicated page <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=Someone%20should%20hat,revenge-porn%20company%20is&amp;text=insufficient.%20Wikipedia%20is,Hold%20the%20users\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">Someone should hat this nonsense part of the thread. \\~2026-47637-7 (talk) 16:41, 22 January 2026 (UTC)\\[\\] To clarify, my argument was one of optics. After Grokipedia, after the massive controversy just recently, the already bad headline \"An organization is paying people to flood Wikipedia with AI content\" becomes the even worse \"An organization is paying people to flood Wikipedia with Grok content.\" It's just an unforced PR error. Gnomingstuff (talk) 02:01, 23 January 2026 (UTC)\\[\\] And ...</span></span></sup>.</td>\n</tr>\n</tbody>\n</table>\n<h3>Overall alignment</h3>\n<p>Broadly, editors fall into three camps:</p>\n<ol>\n<li><strong>Restrictive</strong> (qcne, ClaudineChionh, Iseult, Pi.1415926535, Gurkubondinn, Gnomingstuff, Stepwise Continuous Dysfunction, Rosguill, LuniZunie, Hex, 2026-43687-3): Want OKA prohibited from direct editing, or consider asilvering's proposal insufficient; deeply concerned about AI use, labour practices, or both.</li>\n<li><strong>Reformist / conditional support</strong> (asilvering, Chaotic Enby, Ethmostigmus, Blue Sonnet, voorts, NicheSports, Johnbod, CMD, Toadspike, Buffs, Suriname0): Support asilvering's proposal (or something close to it) as a pragmatic, policy-grounded starting point, while acknowledging it may not be enough on its own.</li>\n<li><strong>Sympathetic to OKA / status-quo-leaning</strong> (7804j, Piotrus, Czarking0, WhatamIdoing, Nihil novi, Toadspike in part): Argue the project is a net positive, errors are marginal, and proposed restrictions risk being disproportionate or unfair; push back on the exploitation framing and on anti-LLM sentiment.</li>\n</ol>\n<p>No binding consensus has been reached. There is broad agreement that <em>something</em> needs to change around verification and oversight, but sharp disagreement on scope (OKA-specific vs. all translations), severity (blocks vs. bans vs. process tweaks), and whether the fundamental model of paying people to add LLM-assisted content is acceptable at all <sup class=\"trigger\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\" target=\"_blank\">1</a><span class=\"tooltip top center citation\"><a href=\"https://en.wikipedia.org/wiki/Wikipedia:Administrators'_noticeboard#OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of_articles.#:~:text=%23%23%20OKA%3A%20problematic,should%20look%20different&amp;text=I%20would%20not,the%20English%20Wikipedia&amp;text=articles%20like%20that,an%20optional%20backlog&amp;text=Articles%20of%20this,this%20comment%20%22%5C%5B&amp;text=%5C%5D%20we%20may,Wikipedia%20languages%20%5C%5B&amp;text=%5C%5D%22%20shows%20that,January%202026%20%28UTC%29%5C%5B%5C%5D&amp;text=yes%2C%20and%3F%207804j,History%20of%20France&amp;text=I%27m%20really%20not,extra%20sensationalism%20here&amp;text=7804j%20%28talk%29%2013%3A43%2C,argumentum%20ad%20hitlerum&amp;text=Piotr%20Konieczny%20aka,web%20portal%20featuring&amp;text=softcore%20pornography%22.%20How,a%20bit%20different&amp;text=_%E2%80%93%20LuniZunie_%28talk%29%2014%3A57%2C,a%20bit%20different&amp;text=Piotr%20Konieczny%20aka,January%202026%20%28UTC%29%5C%5B%5C%5D\"><span class=\"source\"><img alt=\"\" src=\"https://p.kagi.com/proxy/favicons?c=lQ30-cxV_E2Vdyqb6hrbuYgId7qu7hecjO7o1T7aqQDFc7XpcLttL1Ryala-GNHJuWyqBKj8MT4ZBZYnSH4uzuNLNQdbiyz2-gUDYI6hGs0%3D\"> en.wikipedia.org</span><b>1. Wikipedia:Administrators' noticeboard - Wikipedia</b></a><span class=\"passage\">## OKA: problematic paid translation and lead rewrites via LLMs across thousands of articles. \\[edit\\] ### Context \\[edit\\] * OKA is a Swiss non-profit that provides monthly stipends to full-time contributors and translators and leverage AI (Large Language Models) to automate most of the work. OKA editors are required to make a PAID declaration. * OKA’s metawiki Instructions for editors describe extensive machine/LLM-based translation workflows including to Leverage Grok as a starting point. ...</span></span></sup>.</p>","md":"<details><summary>Thinking</summary>\n\nThe user wants me to identify every editor who contributed to this AN discussion and summarise their positions. Let me go through the thread carefully and compile this.\n\n</details>\n\n## Editors and their positions in the OKA AN discussion\n\nBelow is a table of every identifiable editor who contributed substantively to the thread, along with a summary of their position. Editors are listed roughly in the order they first appear.\n\n| Editor | Position summary |\n|---|---|\n| **Qcne** | Thread opener. Documented the long history of problems with OKA translations and proposed remedies (incremental edits, on-wiki consensus for campaigns, transparency on QA). Grew increasingly critical of 7804j's evasive and LLM-drafted replies. Ultimately called for OKA editors to be **prohibited from directly editing articles** [^1]. |\n| **Johnbod** | Long-standing critic of OKA quality. Described the problem as \"long-running and serious\", noted 7804j's intermittent engagement and poor QC. Sceptical that proposed remedies would work with the current set-up, but ultimately **supported** asilvering's proposal as \"a decent start\" [^1]. |\n| **DoubleGrazing** | Noted they had accepted some OKA drafts at AfC that seemed fine, but acknowledged they hadn't specifically looked for MT problems and questioned whether detecting them was within AfC's remit [^1]. |\n| **WhatamIdoing** | Pushed back on the assumption that translators can't work in multiple language pairs; noted multilingualism is normal in many parts of the world. Also contributed a lengthy (and controversial) aside on CSAM statistics [^1]. |\n| **Clovermoss** | Briefly noted that language fluency is a spectrum and that limited fluency can restrict translation direction [^1]. |\n| **Cremastra** | Agreed with the fluency-is-a-spectrum point. Later made a sardonic remark about AI proponents not doing cleanup work [^1]. |\n| **CMD** | Flagged the incident where 7804j detagged obviously LLM-generated text with fabricated sources. Supported asilvering's proposal point 1 (as a general principle for all translations) and point 3 [^1]. |\n| **7804j** (OKA founder) | Defended OKA throughout. Argued errors are statistical outliers in a large project; described QA processes (warnings, grant termination); disclosed stipend amounts and organisational structure. Acknowledged using LLMs to draft his own discussion replies. Receptive to a peer-review pilot and some process changes, but pushed back on characterisations of exploitation and on requirements to verify offline sources. Dismissed the thread as unrepresentative of broader community sentiment [^1]. |\n| **ExtantRotations** | Highlighted that OKA's training guide instructs translators to create multiple email accounts to circumvent LLM usage caps. Questioned the \"grant not employment\" framing and quoted OKA's own website about maximising impact by recruiting in low-cost-of-living countries [^1]. |\n| **ClaudineChionh** | Asked detailed, structured questions about onboarding, selection, supervision, and training. Criticised 7804j for not seeking community approval before launching OKA. Expressed disgust at the gig-work arrangement and power dynamics, but gave **weakest possible support** to asilvering's proposal out of concern for OKA workers [^1]. |\n| **Iseult** | Raised the general-to-native-language translation norm. Concerned about the scale mismatch (few volunteers policing many paid editors). Called for a **three-strikes-and-out** system and gave a **moral oppose** to asilvering's proposal as insufficient [^1]. |\n| **Gurkubondinn** | Supported prohibiting OKA from direct article editing. Questioned compensation practices, asked 7804j to stop using LLM-drafted replies, and pressed for clarity on onboarding payment terms [^1]. |\n| **Chaotic Enby** | Probed onboarding details and labour conditions (independent contractor status, low pay). Conducted **spot checks** of French-translated articles and found errors (hallucinated sources, unsourced expansions). Supported asilvering's proposal plus additional transparency measures (on-wiki consensus for mass campaigns). Later struck concern about targeted donations after 7804j clarified none had been received [^1]. |\n| **Nickoka** | Mentioned by 7804j as co-manager of OKA (handles operations and finances). Not an active on-wiki editor; did not post in the thread [^1]. |\n| **Piotr Konieczny aka Prokonsul Piotrus** | Most prolific defender of OKA alongside 7804j. Argued the project is a net positive; errors are marginal; accused critics of fearmongering and AGF violations. Opposed requiring verification of offline sources as excessive. **Supported** asilvering's proposal points 2–4 but **opposed** point 1 as ambiguously worded. Defended LLM use in drafting talk-page replies and pushed back strongly on the Grok backlash [^1]. |\n| **Gnomingstuff** | Raised the PR/optics risk of using Grok specifically. Conducted a spot check of the Polish article *Walecznych Street* and found several meaning-changing errors. Gave a **moral oppose** to asilvering's proposal, arguing the fundamental problem is paying people to add AI content at all [^1]. |\n| **Ethmostigmus** | Offered a nuanced view: acknowledged OKA's value but stressed the need for better internal oversight and transparency, especially around paid-editing disclosure and targeted donations. Gave **weak support** to asilvering's proposal as a bare minimum [^1]. |\n| **Hex** | Criticised OKA's use of Grok, calling it a \"CSAM generator\" and questioning the judgement behind using it [^1]. |\n| **LuniZunie** | Strongly critical of 7804j's AI-generated replies and lack of respect for community wishes. Wanted more proactive, pre-emptive community engagement from OKA rather than after-the-fact discussions [^1]. |\n| **asilvering** | Author of the **formal four-point proposal** (verify all additions; four-warning block threshold within six months; presumptive deletion of blocked editors' unverified content; prospective application only). Defended the verification requirement against objections about offline sources [^1]. |\n| **Suriname0** | Supportive of asilvering's proposal but suggested point 1 might be slightly too strict for offline sources; proposed a \"reasonable plausibility\" standard for citations that can't be directly checked [^1]. |\n| **voorts** | **Supported** asilvering's proposal. Argued every editor is responsible for WP:V compliance; technically suggested it should be a five-strike rule. Also pushed back on anti-LLM sentiment as out of step with real-world trends [^1]. |\n| **Toadspike** | Investigated OKA stipend levels against World Bank income data and concluded the exploitation charge was not well-founded. Gave **moral support** to asilvering's proposal but questioned whether it would actually change enforcement in practice [^1]. |\n| **Pi.1415926535** | Firmly agreed OKA editors should be prohibited from direct article editing; criticised 7804j for launching OKA without community input and for not addressing major concerns [^1]. |\n| **Stepwise Continuous Dysfunction** | Agreed with prohibiting direct editing. Later **opposed** asilvering's proposal, demanding OKA be barred entirely until it drops all use/endorsement of Grok [^1]. |\n| **Blue Sonnet** | Asked practical questions about escalation contacts, ongoing QA, and post-probation oversight. **Supported** asilvering's proposal, emphasising that paid editors with extra training should be held to a *higher* standard [^1]. |\n| **NicheSports** | **Supported** asilvering's proposal as providing day-one accountability and lowering the barrier to enforcement compared to a full ANI case [^1]. |\n| **Rosguill** | **Moral oppose** to asilvering's proposal: said they would have blocked 7804j on the spot for CIR/NOTHERE-type behaviour; concerned about exploiting income differentials and the dynamics of volunteers policing paid workers. Would prefer OKA to go back to the drawing board [^1]. |\n| **Nihil novi** | Asked why human translators aren't used instead of LLMs. Agreed translators shouldn't have to verify original-article sourcing. **Opposed** asilvering's proposal on the ground that source reliability is the original author's responsibility, not the translator's [^1]. |\n| **Oliwiasocz** (OKA translator) | Responded to the spot-check critique of the *Walecznych Street* article, explaining and correcting the flagged issues [^1]. |\n| **Buffs** | **Supported if necessary**: saw no problem with openly disclosed paid editing by reasonable actors, but questioned whether the proposal was needed [^1]. |\n| **Czarking0** | Defended OKA against exploitation charges; argued the stipend criticism was paternalistic. An AI proponent who patrols NPP; accused critics of confirmation bias. Generally supportive of OKA continuing [^1]. |\n| **2026-43687-3** (IP/temp account) | Called for OKA to be **banned outright**, characterising it as exploitative and neo-colonialist [^1]. |\n| **Wakelamp** | Raised WMF governance and grant-oversight questions (circular funding, eligibility criteria, voting rights of paid editors). Proposed **moving the discussion to a dedicated Wikipedia-space page** and offered to create a structured summary using AI tools [^1]. |\n| **Sesquilinear** | Noted that people can consistently both use LLMs for information and oppose LLM-generated content on Wikipedia, since Wikipedia is a key *input* to LLMs [^1]. |\n| **Daveosaurus** | Provided a Google Street View check of the *Walecznych Street* building, suggesting \"glazed stairwell\" was in fact correct [^1]. |\n| **Donald Albury** | Linked to the *Glazing (window)* article to clarify terminology [^1]. |\n| Various IPs/temp accounts (2026-47637-7, 2026-51814-4, 2025-38536-45, 2026-54906-2, 2025-31850-11, 2026-92659-0, 2026-10381-58) | Brief interjections: some called for hatting tangential sub-threads, some questioned the direction of the discussion, one supported moving it to a dedicated page [^1]. |\n\n### Overall alignment\n\nBroadly, editors fall into three camps:\n\n1. **Restrictive** (qcne, ClaudineChionh, Iseult, Pi.1415926535, Gurkubondinn, Gnomingstuff, Stepwise Continuous Dysfunction, Rosguill, LuniZunie, Hex, 2026-43687-3): Want OKA prohibited from direct editing, or consider asilvering's proposal insufficient; deeply concerned about AI use, labour practices, or both.\n2. **Reformist / conditional support** (asilvering, Chaotic Enby, Ethmostigmus, Blue Sonnet, voorts, NicheSports, Johnbod, CMD, Toadspike, Buffs, Suriname0): Support asilvering's proposal (or something close to it) as a pragmatic, policy-grounded starting point, while acknowledging it may not be enough on its own.\n3. **Sympathetic to OKA / status-quo-leaning** (7804j, Piotrus, Czarking0, WhatamIdoing, Nihil novi, Toadspike in part): Argue the project is a net positive, errors are marginal, and proposed restrictions risk being disproportionate or unfair; push back on the exploitation framing and on anti-LLM sentiment.\n\nNo binding consensus has been reached. There is broad agreement that *something* needs to change around verification and oversight, but sharp disagreement on scope (OKA-specific vs. all translations), severity (blocks vs. bans vs. process tweaks), and whether the fundamental model of paying people to add LLM-assisted content is acceptable at all [^1].","profile":{"id":"1b6d99f4-1649-48c8-8ee3-d6774318f814","name":"wikimedia-policy","accessible":true,"model":"claude-4-6-opus-thinking","deprecate":false,"retired":false,"successor_model":null,"model_name":"Claude 4.6 Opus (reasoning)","model_provider":"anthropic","model_input_limit":1000000,"recommended":true,"model_info":"\n\n\n\n<div class=\"heading\">\n <div class=\"left\">\n \n \n \n <svg class=\"model-icon icon-xs\" data-model=\"claude-4-6-opus-thinking\">\n <use href=\"#icon-sparkles\"/>\n </svg>\n <h3>\n \n <span class=\"model-name\">wikimedia-policy</span>\n </h3>\n </div>\n\n \n <a href=\"/settings?p=custom_assistant&id=1b6d99f4-1649-48c8-8ee3-d6774318f814\">\n <svg class=\"icon-sm edit-icon\" aria-hidden=\"true\">\n <use href=\"#icon-pencil\" />\n </svg>\n </a>\n \n</div>\n\n\n<hr />\n<ul>\n \n <li>\n <h4>\n Model\n </h4>\n <span class=\"flex align-center gap-8\">\n <svg class=\"icon-xs icon\">\n <use href=\"#icon-anthropic\" />\n </svg>\n <span class=\"model-name\">Claude 4.6 Opus (reasoning)</span>\n </span>\n </li>\n <li>\n <h4>\n Web access\n </h4>\n <span>On</span>\n </li>\n <li>\n <h4>\n Lenses\n </h4>\n <span></span>\n </li>\n \n</ul>\n\n\n","model_provider_name":"Anthropic","internet_access":true,"personalizations":false,"lens":{"id":21245,"name":null},"shortcut":null,"is_default_profile":false},"metadata":"<li>\n <span class=\"attribute\">Model</span>\n <span class=\"value\"><i class=\"icon-xs mr-4\"><svg><use href=\"#icon-anthropic\"></use></svg></i>Claude 4.6 Opus (reasoning)</span>\n</li>\n<li>\n <span class=\"attribute\">Version</span>\n <span class=\"value\">claude-opus-4-6</span>\n</li>\n<li>\n <span class=\"attribute\">Speed (tok/s)</span>\n <span class=\"value\">41</span>\n</li>\n<li>\n <span class=\"attribute\">Tokens</span>\n <span class=\"value\">43,899</span>\n</li>\n<li>\n <span class=\"attribute\">Cost / Total ($)</span>\n <span class=\"value\">0.36 / 0.41</span>\n</li>\n<li>\n <span class=\"attribute\">End to end time (s)</span>\n <span class=\"value\">91.06</span>\n</li>\n<li>\n <span class=\"attribute\">Submitted</span>\n <span class=\"value\">2026-02-21 18:56</span>\n</li>\n\n","documents":[]}]