--- title: An AI-written article needs no label. Your proofread... description: Anthropic began watermarking all Claude text output on 2 August, applying the mark even to grammar fixes and light editing that the EU AI Act explicitly... canonical: https://daily.dev/posts/an-ai-written-article-needs-no-label-your-proofread-email-gets-one-cyuan9lat twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: An AI-written article needs no label. Your proofread email gets one | daily.dev og:description: Anthropic began watermarking all Claude text output on 2 August, applying the mark even to grammar fixes and light editing that the EU AI Act explicitly... og:url: https://daily.dev/posts/an-ai-written-article-needs-no-label-your-proofread-email-gets-one-cyuan9lat og:image: https://api.daily.dev/og/posts/cYuAN9LaT.png og:image:alt: An AI-written article needs no label. Your proofread email gets one 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. # An AI-written article needs no label. Your proofread email gets one **[The Next Web](https://daily.dev/sources/tnw)** · 6 min read · 0 upvotes · 0 comments ## Summary Anthropic began watermarking all Claude text output on 2 August, applying the mark even to grammar fixes and light editing that the EU AI Act explicitly exempts from disclosure requirements. The watermark signals only that content 'may have been processed' by Claude, not that Claude authored it, and Anthropic has not released the algorithm, though it promises a free detection API. Meanwhile, Article 50's actual labeling requirement is narrow: fully AI-written novels or marketing copy need no label, and even AI-written news content escapes labeling if a named human editor reviewed it. Critics including Bill Gurley and Steven Sinofsky raised concerns about detection monopoly and data retention, while others like Donn Felker argued marking helps prevent models training on their own synthetic output. The watermark is trivially removable by paraphrasing through a second model, and OpenAI and Google have not yet shipped equivalent text marking, despite facing the same 2 December compliance deadline. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Does Anthropic watermark Claude's output even when it's just proofreading or grammar correction? Yes, Claude watermarks any text it processed, including simple grammar fixes and proofreading, even though the EU AI Act specifically exempts standard editing assistance from disclosure requirements. The watermark works at the model output level and cannot distinguish a full AI-written draft from a single corrected comma, so it marks everything Claude touches, not just what it authored. _Developers weighing AI-assisted writing tools against compliance risk can track policy shifts like this on daily.dev._ ### What does a detected Claude watermark actually indicate about a piece of text? A detected mark only means the text may have been processed by Claude, not that Claude wrote it. Anthropic states this directly: asking the model to proofread, translate, or summarize your own writing still produces watermarked output. Conversely, absence of a mark does not prove there was no AI involvement, since short passages, heavy paraphrasing, older models, or stripped metadata can all evade detection. _Anyone deciding how much trust to place in AI content signals can follow this kind of nuance on daily.dev._ ### When does the EU AI Act require labeling AI-generated content for readers, and does an AI-written article always need one? No, Article 50 only requires a label when AI-generated text informs the public on matters of public interest, and even then only if no named, accountable human editor reviewed it. Fully AI-written novels or marketing copy need no label at all. Existing models have until 2 December to comply, with penalties reaching €15 million or 3% of worldwide annual turnover. _Teams building AI publishing workflows in the EU can keep up with compliance deadlines like this via 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: [#claude](https://daily.dev/tags/claude), [#anthropic](https://daily.dev/tags/anthropic), [#ai-governance](https://daily.dev/tags/ai-governance), [#ai-regulation](https://daily.dev/tags/ai-regulation) [View this post on daily.dev](https://daily.dev/posts/an-ai-written-article-needs-no-label-your-proofread-email-gets-one-cyuan9lat) ```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":"An AI-written article needs no label. Your proofread email gets one","url":"https://daily.dev/posts/an-ai-written-article-needs-no-label-your-proofread-email-gets-one-cyuan9lat","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/an-ai-written-article-needs-no-label-your-proofread-email-gets-one-cyuan9lat"},"datePublished":"2026-08-13T14:46:41.318Z","dateModified":"2026-08-13T14:47:42.242Z","description":"Anthropic began watermarking all Claude text output on 2 August, applying the mark even to grammar fixes and light editing that the EU AI Act explicitly...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/73f9cc911e51000555ae3ecf6969ba3d?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/73f9cc911e51000555ae3ecf6969ba3d?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"The Next Web","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":"The Next Web","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/tnw","url":"https://daily.dev/sources/tnw"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/an-ai-written-article-needs-no-label-your-proofread-email-gets-one-cyuan9lat","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"claude,anthropic,ai-governance,ai-regulation","timeRequired":"PT6M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The Next Web","item":"https://daily.dev/sources/tnw"},{"@type":"ListItem","position":3,"name":"An AI-written article needs no label. Your proofread email gets one"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/an-ai-written-article-needs-no-label-your-proofread-email-gets-one-cyuan9lat#faq","mainEntity":[{"@type":"Question","name":"Does Anthropic watermark Claude's output even when it's just proofreading or grammar correction?","acceptedAnswer":{"@type":"Answer","text":"Yes, Claude watermarks any text it processed, including simple grammar fixes and proofreading, even though the EU AI Act specifically exempts standard editing assistance from disclosure requirements. The watermark works at the model output level and cannot distinguish a full AI-written draft from a single corrected comma, so it marks everything Claude touches, not just what it authored. Developers weighing AI-assisted writing tools against compliance risk can track policy shifts like this on daily.dev."}},{"@type":"Question","name":"What does a detected Claude watermark actually indicate about a piece of text?","acceptedAnswer":{"@type":"Answer","text":"A detected mark only means the text may have been processed by Claude, not that Claude wrote it. Anthropic states this directly: asking the model to proofread, translate, or summarize your own writing still produces watermarked output. Conversely, absence of a mark does not prove there was no AI involvement, since short passages, heavy paraphrasing, older models, or stripped metadata can all evade detection. Anyone deciding how much trust to place in AI content signals can follow this kind of nuance on daily.dev."}},{"@type":"Question","name":"When does the EU AI Act require labeling AI-generated content for readers, and does an AI-written article always need one?","acceptedAnswer":{"@type":"Answer","text":"No, Article 50 only requires a label when AI-generated text informs the public on matters of public interest, and even then only if no named, accountable human editor reviewed it. Fully AI-written novels or marketing copy need no label at all. Existing models have until 2 December to comply, with penalties reaching €15 million or 3% of worldwide annual turnover. Teams building AI publishing workflows in the EU can keep up with compliance deadlines like this via daily.dev."}}]} ```