Market & Regulation
Anthropic Details Claude's Invisible Text Watermarks
The Verge AIhigh confidence
Why it matters
- →Ensures compliance with the EU's AI transparency rules for synthetic content.
- →Uses open-source technology to create detectable patterns without affecting output quality.
- →Sets a precedent for other AI developers facing similar regulatory requirements.
Anthropic has announced plans to integrate invisible text watermarks into its Claude AI model, utilizing a version of Google DeepMind's SynthID-Text system. This technology creates detectable patterns in text through wording probabilities, ensuring compliance with the European Union's AI transparency rules. The watermarks are designed to be invisible to users but detectable by those with the necessary key, allowing for compliance without impacting the quality or cost of Claude's outputs. This initiative is part of a broader industry response to the EU's AI Act requirements.
Read original
