Anthropic announced on November 3 that text from Claude Mythos 5 (claude-mythos-5), Claude Opus 4.7 (claude-opus-4-7), and Claude Sonnet 4.6 (claude-sonnet-4-6) now carries its embedded text watermark. The change covers supported first-party and cloud platforms worldwide, including API output. Amazon Bedrock rollout for these models is due to finish by November 10.
Anthropic says the watermark adds no tokens or hidden characters and does not change meaning or readability. Sonnet 4.6 outputs may lack it when temperature, top_p, or top_k uses a non-default value. Detection remains in private preview for eligible organizations, and a detected mark is not conclusive proof of authorship. The change expands traceability for existing models, with practical limits on who can verify the signal.
Source · platform.claude.com ↗
