C2PA

Anthropic confirmed it will watermark AI-generated text and files from Claude models, complying with the EU AI Act's Transparency Code that took effect August 2. The watermark is applied at the model level and persists through copy-paste, according to the company.