Claude text now carries Anthropic's text watermark on three models (November 3, 2026)

What changed

On November 3, 2026 Anthropic's release notes said that text from Claude Mythos 5, Claude Opus 4.7 and Claude Sonnet 4.6 carries Anthropic's text watermark.

The docs say the watermark does not change the meaning, quality or readability of the output. It adds no tokens or hidden characters and carries no information about you or your organization. You do not need to change your requests or how you handle responses.

On Claude Sonnet 4.6, text from a request that sets temperature, top_p or top_k to a non-default value might not carry the watermark.

Prerequisites

• An Anthropic API key in ANTHROPIC_API_KEY
• An app that calls Claude Mythos 5, Claude Opus 4.7 or Claude Sonnet 4.6

Steps

1. Nothing to change if you only want normal output.
2. If you rely on the watermark and call Claude Sonnet 4.6, remove temperature, top_p and top_k from your requests, including any that a framework or client library sets for you.
3. For the full list of models and platforms it covers, read Anthropic's help article "How Claude marks AI-generated content", linked from the text watermark page.

Test it

Send the same request to Claude Sonnet 4.6 before and after removing the sampling parameters. Both should return a normal reply; the docs say the watermark adds no tokens, so token counts should not change because of it.

Not confirmed

• The docs page shows no API or tool for checking whether a piece of text carries the watermark.
• The docs page does not list the covered models itself; it points to a help center article.
• No code sample: the official pages have none.