Anthropic has signed the EU AI Act Article 50(2) Code of Practice on Transparency of AI-Generated Content as a provider of generative AI models and systems. This article describes how Anthropic plans to implement those commitments through machine-readable marks in Claude-generated content.

Claude uses two complementary marking techniques: (1) imperceptible watermarks embedded directly into generated text that persist through copying and some editing, applied at the model level across all Claude products; and (2) signed provenance metadata attached to supported file types (.svg, .png, .jpg) following the C2PA open standard, which signals Claude processing and detects tampering.

Anthropic is working to enable users and third parties to detect these marks through forthcoming technical documentation. Detection checks whether content carries a supported Claude mark, indicating potential Claude processing.

Machine-readable marks have limitations across content types. Builders deploying Claude should independently assess Article 50 requirements for their products, and Anthropic will share technical guidance to support transparency obligations.