Checked 4 November 2026.
Claude extends text-watermark coverage to three earlier models
Anthropic’s 3 November release notes say text from Claude Mythos 5, Opus 4.7 and Sonnet 4.6 now carries its watermark on supported platforms. This extends coverage to earlier models; it is not the first launch of Claude watermarking. Original release notes.
The coverage guide says marking applies worldwide where supported Claude models are offered, including Claude products and cloud partners. For these three models, Amazon Bedrock rollout is due to finish by 10 November. Text-watermark detection remains a private preview for eligible organisations, rather than a generally available public checker.
Tecrave analysis: A small agency producing client copy should check which models its writing tools actually use and keep a clear record of AI assistance. Proofreading, translation or summarisation can also produce marked text, so a mark alone should not be treated as proof that the original ideas or entire document came from AI.
Anthropic says its text watermark adds no tokens or hidden characters and requires no change to ordinary request handling. Its technical guide warns that Sonnet 4.6 outputs may lack the mark when sampling settings are changed from their defaults. The help guide also cautions that an absent mark does not prove a passage was written without AI. Tecrave has not independently tested detection or watermark durability.
Prepared with AI assistance from linked primary sources.