Anthropic announced plans to embed imperceptible watermarks in text generated by its Claude models in order to comply with EU AI Act transparency requirements. The company said generated text will carry embedded watermarks and that generated files will include digitally signed provenance metadata where supported, using the C2PA standard for files. The changes will apply to Claude outputs globally, across its consumer app, API, Claude Code, Claude Cowork, and versions of Claude accessed through AWS, Google Cloud, and Microsoft Foundry. Some researchers and users have questioned the effectiveness of text watermarking, noting that comparable image-watermarking schemes have previously been circumvented.
