Anthropic plans to add text watermarking to Claude Opus 4.6, Opus 4.5, Sonnet 4.5 and Haiku 4.5 starting November 16, according to an email to API customers. The rollout on Amazon Bedrock may take a few extra days. Anthropic gave no reason for delaying Opus 4.6, which had originally been scheduled to receive the watermark on November 3.
For all four models in this rollout, the watermark applies only when temperature, top_p and top_k are set to their default values. Changing any one of those sampling parameters can bypass the mark. Anthropic advised customers who depend on the watermark for compliance to leave the settings unchanged.
Anthropic describes its watermark as an imperceptible pattern in word choice that does not add characters or tokens. The company says it does not change meaning, quality or readability, and that it contains no information about users or their conversations. A detection API for checking the watermark is in private preview.
The company says it is retrofitting older models ahead of the EU AI Act’s December deadline. Models released after August 2, 2026, launched with watermarking enabled, while earlier models are receiving it in stages. Anthropic says the mark is embedded at the model level and cannot be turned off by customers on the platforms hosting the models.
Sources: The Decoder.