Anthropic to embed invisible watermarks in Claude's text

Anthropic said on August 11 it will add an imperceptible, machine-readable watermark to text generated by new Claude models starting August 2, responding to transparency requirements under Article 50 of the EU AI Act. The mark stays in the text through copy and paste, applies worldwide rather than only in Europe, and reaches Claude wherever it is served, including the API, the consumer app, Claude Code, Claude Cowork and Claude Tag, and through AWS, Google Cloud and Microsoft Foundry; generated files separately get signed provenance metadata under the C2PA standard, and Anthropic said it is working to extend marking to older models during the AI Act's transition period.