--- title: AI 'watermark removers' flood the web. Almost none can... description: Days after Anthropic began invisibly watermarking Claude-generated text to comply with EU AI Act Article 50, a wave of 'watermark remover' tools has appeared,... canonical: https://daily.dev/posts/ai-watermark-removers-flood-the-web-almost-none-can-prove-they-work--iypcksxfu twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI 'watermark removers' flood the web. Almost none can prove they work. | daily.dev og:description: Days after Anthropic began invisibly watermarking Claude-generated text to comply with EU AI Act Article 50, a wave of 'watermark remover' tools has appeared,... og:url: https://daily.dev/posts/ai-watermark-removers-flood-the-web-almost-none-can-prove-they-work--iypcksxfu og:image: https://api.daily.dev/og/posts/IypcKSxfU.png og:image:alt: AI 'watermark removers' flood the web. Almost none can prove they work. 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. # AI 'watermark removers' flood the web. Almost none can prove they work. **[BleepingComputer](https://daily.dev/sources/bleepingcomputer)** · 5 min read · 0 upvotes · 0 comments ## Summary Days after Anthropic began invisibly watermarking Claude-generated text to comply with EU AI Act Article 50, a wave of 'watermark remover' tools has appeared, including an open source project with over 4,500 GitHub stars and paid AI-detection-evasion services like StealthGPT and Human Writes. None of their claims can be verified because Anthropic has not released a public detector or technical documentation. The tools reliably strip hidden Unicode characters and file metadata (C2PA/EXIF/XMP), but the actual text watermark lives in word choice and can only be removed by heavily rewriting content with a second model - something the leading tool's own author admits doesn't work yet. Independent code review found one popular cleaner failed to strip a common hidden-payload technique. The bigger concern raised is supply-chain risk: these tools are being wired directly into agent pipelines and fed sensitive documents, with unaudited code and a growing market with no way to verify vendor claims. ## 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 text generated by Claude now? Yes, text from Claude models launched on or after August 2, 2026 carries an imperceptible watermark woven into the wording itself, based on which words the model chooses. Supported file types also get signed C2PA metadata. Marking applies across the API, claude.ai, Claude Code, Claude Cowork, Claude Tag, and through AWS, Google Cloud and Microsoft Foundry, driven by EU AI Act Article 50 compliance. _Developers building on Claude can track provenance and compliance changes like this via daily.dev._ ### Can existing watermark remover tools actually strip Anthropic's Claude text watermark? No verified tool can strip it yet. Because the watermark lives in the model's word choices rather than hidden characters, removing it requires heavily rewriting text with a second model, which the most prominent tool's own creator admits isn't implemented yet. Existing tools reliably strip hidden Unicode characters and file metadata (C2PA, EXIF, XMP), but that doesn't touch the actual watermark, and no public detector exists to verify any claim. _Anyone evaluating AI content provenance tools can follow how these claims hold up on daily.dev._ ### Why did Anthropic start watermarking Claude-generated content? The trigger is Article 50 of the EU AI Act, which became enforceable on August 2, 2026 and carries penalties of up to 15 million euros or 3% of global turnover for non-compliance. A detected watermark only indicates content was processed by Claude, not necessarily authored by it, and Anthropic itself notes heavy editing, paraphrasing, or translation can remove the mark. _Teams navigating EU AI Act compliance can keep tabs on rules like Article 50 through daily.dev._ --- Tags: [#security](https://daily.dev/tags/security), [#claude](https://daily.dev/tags/claude), [#anthropic](https://daily.dev/tags/anthropic) [View this post on daily.dev](https://daily.dev/posts/ai-watermark-removers-flood-the-web-almost-none-can-prove-they-work--iypcksxfu) ```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":"AI 'watermark removers' flood the web. Almost none can prove they work.","url":"https://daily.dev/posts/ai-watermark-removers-flood-the-web-almost-none-can-prove-they-work--iypcksxfu","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-watermark-removers-flood-the-web-almost-none-can-prove-they-work--iypcksxfu"},"datePublished":"2026-08-13T17:35:35.133Z","dateModified":"2026-08-14T22:21:32.821Z","description":"Days after Anthropic began invisibly watermarking Claude-generated text to comply with EU AI Act Article 50, a wave of 'watermark remover' tools has appeared,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/aa05cf1000990b2362b0200a0e58bf09?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/aa05cf1000990b2362b0200a0e58bf09?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"BleepingComputer","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":"BleepingComputer","logo":"https://media.daily.dev/image/upload/s--as8nJ3qy--/f_auto,q_auto/v1774959951/logos/bleepingcomputer?_a=BAMAMiWQ0","url":"https://daily.dev/sources/bleepingcomputer"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-watermark-removers-flood-the-web-almost-none-can-prove-they-work--iypcksxfu","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"security,claude,anthropic","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"BleepingComputer","item":"https://daily.dev/sources/bleepingcomputer"},{"@type":"ListItem","position":3,"name":"AI 'watermark removers' flood the web. Almost none can prove they work."}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-watermark-removers-flood-the-web-almost-none-can-prove-they-work--iypcksxfu#faq","mainEntity":[{"@type":"Question","name":"Does Anthropic watermark text generated by Claude now?","acceptedAnswer":{"@type":"Answer","text":"Yes, text from Claude models launched on or after August 2, 2026 carries an imperceptible watermark woven into the wording itself, based on which words the model chooses. Supported file types also get signed C2PA metadata. Marking applies across the API, claude.ai, Claude Code, Claude Cowork, Claude Tag, and through AWS, Google Cloud and Microsoft Foundry, driven by EU AI Act Article 50 compliance. Developers building on Claude can track provenance and compliance changes like this via daily.dev."}},{"@type":"Question","name":"Can existing watermark remover tools actually strip Anthropic's Claude text watermark?","acceptedAnswer":{"@type":"Answer","text":"No verified tool can strip it yet. Because the watermark lives in the model's word choices rather than hidden characters, removing it requires heavily rewriting text with a second model, which the most prominent tool's own creator admits isn't implemented yet. Existing tools reliably strip hidden Unicode characters and file metadata (C2PA, EXIF, XMP), but that doesn't touch the actual watermark, and no public detector exists to verify any claim. Anyone evaluating AI content provenance tools can follow how these claims hold up on daily.dev."}},{"@type":"Question","name":"Why did Anthropic start watermarking Claude-generated content?","acceptedAnswer":{"@type":"Answer","text":"The trigger is Article 50 of the EU AI Act, which became enforceable on August 2, 2026 and carries penalties of up to 15 million euros or 3% of global turnover for non-compliance. A detected watermark only indicates content was processed by Claude, not necessarily authored by it, and Anthropic itself notes heavy editing, paraphrasing, or translation can remove the mark. Teams navigating EU AI Act compliance can keep tabs on rules like Article 50 through daily.dev."}}]} ```