AI Matters
Why it matters
Anthropic's AI text watermark can be switched off on older Claude models
The Decoder Anthropic's text watermark is easy to bypass on several older Claude models
Anthropic is adding text watermarks to four more Claude models before the EU AI Act deadline, but API users can disable them on older models by changing one setting.
Why it matters
Marks meant to identify AI-written text may be easy to remove, which matters for spotting AI content under new EU rules.
Latest brief · Nov 4, 2026