--- title: Our approach to EU text provenance rules | daily.dev description: OpenAI outlines its phased response to the EU AI Act's text provenance requirements, introducing an invisible text watermarking technology called textGrain.... canonical: https://daily.dev/posts/our-approach-to-eu-text-provenance-rules-6tacjwsfc twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Our approach to EU text provenance rules | daily.dev og:description: OpenAI outlines its phased response to the EU AI Act's text provenance requirements, introducing an invisible text watermarking technology called textGrain.... og:url: https://daily.dev/posts/our-approach-to-eu-text-provenance-rules-6tacjwsfc og:image: https://api.daily.dev/og/posts/6TaCjWSfC.png og:image:alt: Our approach to EU text provenance rules og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Our approach to EU text provenance rules **[OpenAI](https://daily.dev/sources/openai)** · 7 min read · 0 upvotes · 0 comments ## Summary OpenAI outlines its phased response to the EU AI Act's text provenance requirements, introducing an invisible text watermarking technology called textGrain. API customers globally can opt in to watermarking for select models starting now, with ChatGPT and Codex in the EU getting invisible watermarks in coming weeks. Detector access is initially limited to approved researchers and expert organizations due to reliability concerns: detection drops significantly for short text, math content, and edited passages (replacing 25% of words with synonyms cuts detection from 92% to 17%). OpenAI stresses watermarks cannot measure human contribution, establish ownership, identify users, or verify accuracy, and plans to open-source the technology eventually. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How does OpenAI's textGrain text watermarking work and how reliable is it? TextGrain adds an invisible statistical signal to a language model's word choices, which a detector scans for to assess whether a passage contains an OpenAI watermark. At a 1% false positive rate, it detects watermarks in about 80% of 200-token passages and 95% of 400-token passages, though math content performs worse. Replacing 25% of words with synonyms drops detection to just 17%. _Developers weighing AI content compliance can follow watermarking reliability updates like this on daily.dev._ ### Who can access OpenAI's text watermark detector under the EU AI Act rules? Access is initially limited to approved researchers and expert organizations that apply through OpenAI's content provenance form, granted case-by-case in line with the EU's Code of Practice on AI-generated content. It is not publicly available at launch due to the risk of false positives and missed watermarks. The detector only reports whether a watermark is present, without identifying the user or their prompts. _Teams tracking AI compliance requirements can stay current on rollouts like this via daily.dev._ ### When will ChatGPT and the OpenAI API start supporting text watermarking? API customers worldwide can opt in to watermarked text outputs for select models starting immediately, with watermarking off by default. Over the following weeks, an invisible watermark will be added to eligible ChatGPT and Codex text output specifically within the European Union, not as a global default, to let OpenAI gather real-world feedback first. _Developers integrating OpenAI's API can track feature rollouts like this one on daily.dev._ ## Similar posts on daily.dev - [Text AI watermarks will always be trivial to remove](https://daily.dev/posts/text-ai-watermarks-will-always-be-trivial-to-remove-borbz7x9u) · sean goedecke · 6 upvotes · 4 comments --- Tags: [#openai](https://daily.dev/tags/openai), [#chatgpt](https://daily.dev/tags/chatgpt), [#ai-regulation](https://daily.dev/tags/ai-regulation) [View this post on daily.dev](https://daily.dev/posts/our-approach-to-eu-text-provenance-rules-6tacjwsfc) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Our approach to EU text provenance rules","url":"https://daily.dev/posts/our-approach-to-eu-text-provenance-rules-6tacjwsfc","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/our-approach-to-eu-text-provenance-rules-6tacjwsfc"},"datePublished":"2026-10-05T15:18:59.371Z","dateModified":"2026-10-05T21:49:05.215Z","description":"OpenAI outlines its phased response to the EU AI Act's text provenance requirements, introducing an invisible text watermarking technology called textGrain....","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/010b2248d9ba61eeb81df25cdbc50568?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/010b2248d9ba61eeb81df25cdbc50568?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"OpenAI","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"OpenAI","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/984b3632316f4afaa3d5213d5c64fe65","url":"https://daily.dev/sources/openai"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/our-approach-to-eu-text-provenance-rules-6tacjwsfc","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"openai,chatgpt,ai-regulation","timeRequired":"PT7M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"OpenAI","item":"https://daily.dev/sources/openai"},{"@type":"ListItem","position":3,"name":"Our approach to EU text provenance rules"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/our-approach-to-eu-text-provenance-rules-6tacjwsfc#faq","mainEntity":[{"@type":"Question","name":"How does OpenAI's textGrain text watermarking work and how reliable is it?","acceptedAnswer":{"@type":"Answer","text":"TextGrain adds an invisible statistical signal to a language model's word choices, which a detector scans for to assess whether a passage contains an OpenAI watermark. At a 1% false positive rate, it detects watermarks in about 80% of 200-token passages and 95% of 400-token passages, though math content performs worse. Replacing 25% of words with synonyms drops detection to just 17%. Developers weighing AI content compliance can follow watermarking reliability updates like this on daily.dev."}},{"@type":"Question","name":"Who can access OpenAI's text watermark detector under the EU AI Act rules?","acceptedAnswer":{"@type":"Answer","text":"Access is initially limited to approved researchers and expert organizations that apply through OpenAI's content provenance form, granted case-by-case in line with the EU's Code of Practice on AI-generated content. It is not publicly available at launch due to the risk of false positives and missed watermarks. The detector only reports whether a watermark is present, without identifying the user or their prompts. Teams tracking AI compliance requirements can stay current on rollouts like this via daily.dev."}},{"@type":"Question","name":"When will ChatGPT and the OpenAI API start supporting text watermarking?","acceptedAnswer":{"@type":"Answer","text":"API customers worldwide can opt in to watermarked text outputs for select models starting immediately, with watermarking off by default. Over the following weeks, an invisible watermark will be added to eligible ChatGPT and Codex text output specifically within the European Union, not as a global default, to let OpenAI gather real-world feedback first. Developers integrating OpenAI's API can track feature rollouts like this one on daily.dev."}}]} ```