Anthropic Adds Invisible Watermarks to Claude Content

Anthropic has announced plans to embed invisible watermarks in text and images generated by Claude, bringing the AI model into compliance with the EU AI Act's voluntary Code of Practice. Text watermarking applies to Claude models released after August 2, with earlier models to follow. The watermark persists through copying and edits. For images, Anthropic will use C2PA, a metadata-based system that records which model created the content, when, and any copyright details. It includes tamper-detection features. Anthropic also plans to release detection tooling, though it cautioned the system may struggle with short or heavily edited text.