Claude will embed machine-readable watermarks in generated content, including retroactive support for older models, to meet EU transparency rules.
Anthropic said it will apply invisible, machine-readable watermarks to text and images generated by Claude, and will extend the capability to older model versions. The move is aimed at complying with European AI transparency requirements, including C2PA-style provenance standards.
The announcement lands alongside Spotify's separate move to label 'AI Persona' artist profiles and exclude their music from recommendations by default — both signs that platforms are building disclosure infrastructure ahead of regulatory deadlines rather than waiting to be forced.
Watermarking-by-default shifts AI content provenance from a nice-to-have to table stakes for any vendor selling into regulated markets. Operators building on Claude or distributing AI content should expect labeling and disclosure requirements to become a standard procurement checkbox, not an edge case.
The daily signal, curated. Get it in your inbox.
Subscribe on LinkedIn →