As of August 2, texts produced by Claude will incorporate invisible watermarks that are seamlessly integrated into the statistical distribution of the generated content. These watermarks are designed to be undetectable and cannot be eliminated through copying, pasting, or partial editing. Anthropic unveiled this initiative in its help documentation on Monday, aligning with the transparency stipulations outlined in Article 50 of the EU AI Act. This regulation requires providers of generative AI to clearly label their output content, ensuring accountability and traceability.