Claude's new Scarlet Letter watermark is invisible — for now

Anthropic embedded a watermark in outputs from Claude that identifies content processed by the AI system. The watermark is not visible to users in normal viewing but is detectable through technical analysis. According to Anthropic, the watermark flags any text Claude generated or edited, including instances where Claude only modified human-written content. The company stated the watermark persists regardless of whether users attempt to remove it.

Next story