Anthropic expanding text watermark to more Claude models but bypassable on older ones
Original title: Anthropic's text watermark is easy to bypass on several older Claude models
Anthropic is watermarking Claude Opus 4.6, Opus 4.5, Sonnet 4.5, and Haiku 4.5 starting November 16, 2026, but API users can disable it by changing sampling parameters on Opus 4.6 and Sonnet 4.6.
Anthropic is retrofitting watermarking to older Claude models to meet the EU AI Act deadline. The watermark is applied at the model level and affects platforms like Amazon Bedrock. However, on Claude Opus 4.6 and Sonnet 4.6, it only activates when temperature, top_p, and top_k are at default settings. Newer models like Opus 5.5 already have it. Customers cannot disable it.
Unconfirmed: Whether watermarking truly leaves text quality untouched is disputed.
- Watermarking applies to Claude Mythos 5, Opus 4.7, Opus 4.6, and Sonnet 4.6
- Bypass possible on Opus 4.6 and Sonnet 4.6 by tweaking sampling parameters
- Rollout delayed to November 16, 2026 for some models
- Uses variant of Google DeepMind's SynthID
From the original article
3 min read at the-decoder.com
Anthropic is rolling out text watermarks to four more Claude models ahead of the EU AI Act's December deadline. But on older models like Claude Opus 4. 6 and Sonnet 4.
By Matthias Bastian · the-decoder.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.