How Claude marks AI-generated content

Anthropic has committed under the EU AI Act to mark AI-generated content from Claude, implementing machine-readable watermarks in text and digitally signed provenance metadata in files. Starting August 2, 2026, all new Claude models will include these marks at launch across all platforms globally, while existing models will be updated with marking support during a transition period. The company will also provide tools to help users and third parties detect these marks.