As global regulators move from voluntary safety agreements to enforceable transparency frameworks, AI lab Anthropic has signed the European Union’s Code of Practice on Transparency of AI-Generated Content. In accordance with the mandate, the company is implementing comprehensive, machine-readable watermarking across all output generated by its Claude models. While the regulatory requirement originates in the EU, Anthropic is deploying the marking system worldwide across the Claude API, web and mobile apps, Claude Code, Claude Cowork, and Claude Tag. This change also benefits users as the chances of AI detection are significantly lower with cemented watermarking standards.
How does Invisible Text Watermarking work?
Unlike conventional digital documents that embed visible stamps or hidden Unicode characters, Anthropic’s text watermarking operates directly at the model generation level using an adaptation of Google DeepMind’s SynthID-Text methodology.
When large language models generate sentences, they choose each subsequent word from a candidate pool of probable alternatives. In situations where multiple word choices are equally appropriate, a secret cryptographic key combined with the preceding text subtly steers the selection. Across paragraphs of text, these micro-selections form a distinct statistical pattern.
The era of unprovable AI writing just ended.
— Aakash Gupta (@aakashgupta) Starting August 2, every new Claude model weaves an invisible watermark into the words it generates. Copy it, paste it into an email, a blog post, a college essay, and the watermark travels with the text. Anthropic will publish the…[https://t.co/DV4KvyfjCc][August 11, 2026]
The resulting text reads completely naturally to human readers, but a detector equipped with the key can mathematically verify whether the word-choice distribution matches Claude’s generation signature. Anthropic emphasized that this statistical technique adds no extra characters, consumes no additional tokens, has zero impact on output pricing or response quality, and carries no personal user identifiers or chat-specific data.
C2PA cryptographic metadata for file formats
For generated and processed visual file formats–including PNG, JPG, and SVG files–Anthropic uses the open Coalition for Content Provenance and Authenticity (C2PA) industry standard. The system embeds cryptographically signed metadata into the file container indicating that Claude processed or generated the asset. Any C2PA-compatible image inspector or web viewer can read the manifest to confirm provenance and detect whether the file or metadata has been altered after export. Anthropic’s detection capabilities and known limitations
Anthropic noted that watermarking acts as a probabilistic provenance signal rather than absolute proof of authorship. The mark indicates that Claude’s models were involved in generating or processing the wording, but cannot differentiate whether Claude authored the text from scratch or performed heavy proofreading, translation, or structural editing on user-provided drafts.
Anthropic published a full FAQ on how Claude's watermark works. Here is what you need to know. 🤯
— Vaibhav Sisinty (@VaibhavSisinty) → Claude picks between equally good words. The watermark decides which one. You cannot see it. Anyone with the key can.
→ Google tested this on Gemini. Users could not tell the…[https://t.co/6TBd5mTomV][pic.twitter.com/8qQmaRubmz][August 15, 2026] Because the text mark relies on statistical word distribution, minor proofreading and standard copy-pasting will preserve the signature, but heavy rewriting, substantive paraphrasing, or extensive manual editing can dilute the pattern until it becomes undetectable.
All new Claude models released from August 2026 feature watermarking active at launch, with legacy models receiving retroactive support through the EU-permitted transition window. Anthropic is developing a dedicated detection API to allow platforms, enterprises, and educators to programmatically verify Claude-generated text against the cryptographic key. As major AI developers–including Google, Microsoft, and OpenAI–implement similar Code of Practice mechanisms, standardized content provenance is set to become a universal layer across consumer and enterprise generative AI tools.
(Feature image credits to Anthropic.) Read More: Xiaomi's most powerful in-house processor will be featured in its next book-style foldable