Anthropic has committed to embedding machine-readable marks in all content generated by its Claude models, the company said on Tuesday, as it became a signatory to the EU AI Act’s Code of Practice on transparency for AI-generated content. The code, which took effect on August 2, requires providers to label AI-produced material in a way that can be detected by other systems.
According to an official update on Claude's support page, new Claude models released on or after August 2 will include two types of marking from day one. Generated text will carry an imperceptible digital watermark woven directly into the words, which the company says does not affect the quality or readability of the output. Because the watermark is embedded at the model level, it will remain present whether users access Claude through the API, the consumer app, or tools such as Claude Code, Claude Cowork, and Claude Tag. It will also persist when text is copied and pasted elsewhere, and may survive light editing.
For files such as images, Claude will attach signed provenance metadata using the C2PA open standard. This metadata records that the file was processed by Claude and allows detection of tampering, but it can be stripped by format conversion, re-saving, or screenshots.
Anthropic said the markings will be applied globally, not just in the European Union. The company is also working to extend marking support to older models released before the August deadline, though it did not provide a timeline. It plans to release technical documentation to help users and third parties detect its marks.
The company cautioned that a detected mark does not constitute proof of full provenance. For instance, people often use Claude to proofread or translate material they originally created, meaning the underlying ideas may not be AI-generated. Conversely, heavily edited or shortened text, or files with stripped metadata, may not carry a detectable mark.
According to a report by TechCrunch on the matter, Anthropic joins a group of technology companies, including Black Forest Labs, Google, Meta, Microsoft, OpenAI, and Synthesia, that have signed the EU’s transparency code. The broader industry push follows a string of legal challenges and user backlash over AI-generated content posted without disclosure.
