Anthropic is watermarking Claude's output for the EU AI Act

Anthropic has signed the EU AI Act's Code of Practice on Transparency of AI-Generated Content and published how it marks what Claude produces.

The support article describes the commitments it made as a provider of both generative models and generative systems. Claude models launched in the EU on or after 2 August 2026 mark their output from day one: generated text carries embedded watermarks and generated files include Content Credentials, the C2PA provenance standard, where the file type supports it. The marks apply across Claude surfaces including the Claude API, Claude Code and Claude Cowork, and worldwide rather than only in the EU.

Models released before that date are a work in progress. Anthropic says the law includes a transition period for the older models and that it is adding marking support to them.

Detection is not open to everyone yet. Anthropic's watermark detector is in private preview for what it calls eligible organisations as required under EU law, listing regulators, law enforcement, media, fact-checkers, independent researchers, educational organisations and EU civil society groups, plus enterprises that have their own verification duty under the Act. It also says it plans to widen access to the detection API and is taking expressions of interest.

The article carries no fixed date, and the rollout was reported on 25 September 2026 by the provenance newsletter Silent Signals in its daily deepfake risk report. It is the practical half of the EU regime: the Code asks for marking that is effective, interoperable, robust and reliable as far as technically feasible, and Anthropic's answer is two complementary techniques at once: a watermark in the text plus file metadata.

Anthropic's article and the access request form are at support.claude.com; the Code itself is published by the Commission's AI Office.