Anthropic Details Watermarking Strategy for Claude AI Text

Anthropic will implement text watermarking for its Claude chatbot to comply with the EU AI Act’s Transparency Code.

The company plans to use the SynthID-Text approach, which embeds patterns undetectable to readers but identifiable via a specific detection key.

Anthropic stated that while light editing may not remove the watermark, a complete rewrite of the text would effectively eliminate it.