Claude will apply invisible watermarks to AI text and images

Anthropic has committed to embedding machine-readable watermarks and digitally signed provenance metadata into text and images generated by Claude models to comply with the European Union's AI Act transparency rules. The updates will apply globally across various Claude platforms and partner clouds like AWS, Google Cloud, and Microsoft Foundry. While images will utilize the C2PA standard, text will feature an imperceptible embedded watermark that persists through copying and minor edits. Anthropic acknowledges that these systems are not infallible, as metadata can sometimes be stripped, and plans to release further technical documentation on detection tools soon.

(Source：The Verge)