Anthropic watermarks Claude text at the model level to meet the EU AI Act
Anthropic will watermark text and files from every Claude model released after August 2, applying the mark at the model level so it rides every surface — the API, Claude Code, and Claude Cowork included. The text mark travels with copy-and-paste and may persist through some editing, with no stated threshold for stripping it; files use the open C2PA standard. Provenance is moving from a platform feature to a property of the model itself, with Google, Meta, Microsoft, OpenAI, and Synthesia signed on to the same EU code.
Source: techcrunch.com ↗
Because the watermark is part of the text, it will travel with the text when it's copied and pasted elsewhere, and may persist through some editing.
Anthropic support page
Why this matters
- → EU AI Act compliance moves watermarking from optional platform feature to mandatory model-level requirement
- → Watermark persists through copy-paste and editing, making AI provenance traceable across surfaces
- → Seven major AI labs adopt same standard to meet transparency requirements
Regulation reshapes AI provenance