Anthropic extends Claude text watermarking to four models
What happened
The Decoder reported on November 1, 2026 that Anthropic extended text watermarking to Claude Mythos 5, Claude Opus 4.7, Claude Opus 4.6 and Claude Sonnet 4.6, four models in total. The report also notes that wider coverage does not mean every listed model is watermark-proof: two older models, Claude Opus 4.6 and Claude Sonnet 4.6, have a bypass route, and sampling parameters can get around the text watermark. The disclosed progress covers the wider coverage and the bypass on those two older models.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
- The DecoderAnthropic's watermark rollout leaves a loophole on two older Claude models
Anthropic 将文本水印扩展至 Claude Mythos 5、Claude Opus 4.7、Claude Opus 4.6 和 Claude Sonnet 4.6，其中后两款存在绕过入口。
Heat over time
Not enough continuous observations to draw a trend yet.