Text watermark on older Claude models
Text generated by Claude Opus 4.6, Claude Opus 4.5, Claude Sonnet 4.5, and Claude Haiku 4.5 now carries Anthropic's text watermark on supported platforms. The watermark does not change the meaning, quality, or readability of the output, adds no tokens or hidden characters, and carries no information about you or your organization. If a request to one of these models sets temperature, top_p, or top_k to a non-default value, the text might not carry the watermark. Anthropic has signed the EU AI Act's Article 50(2) Code of Practice and says Claude models launched in the EU on or after August 2, 2026 support machine-readable marking at launch. Watermark detection is in private preview for eligible organizations.
Soft Pyramid is not affiliated with, endorsed by, or a partner of Anthropic. These guides are Soft Pyramid's own notes on Claude features.
How to use it
- Check which models and platforms are covered in the table on How Claude marks AI-generated content. Anthropic notes that rollout on Amazon Bedrock may take a few additional days for these models.
- If you need your output to carry the watermark, omit temperature, top_p, and top_k from your requests.
- Check your framework or client library and the request body it sends, and remove any of the three parameters it sets.
- To check marks on your own outputs, register interest through the Claude Watermark Detector Access Request Form linked from the help center article.
