4 outlets·4 reports
Anthropic shares more details about how Claude’s new watermarks will work
ThinkingNews Desk · how this was written
Anthropic’s AI model Claude embeds a text watermark that indicates the output was likely generated by Claude. The watermark is sparse, appearing only in code and factual passages, and it vanishes when the text is completely rewritten. This feature has been implemented to meet European Union regulatory requirements.
Written from all 4 reports below, not from any single one.
How it was reported
- Hacker News·How Claude's text watermarking works
- TechMeme·Anthropic details Claude's text watermark: it only shows Claude was likely involved, is sparse in code and factual text, and disappears after a full rewrite (Anthropic)
- Engadget·Anthropic is watermarking text generated by Claude to comply with EU law
- TechCrunch·Anthropic shares more details about how Claude’s new watermarks will work