★ Top story · Models2d ago
Anthropic watermarks Claude's output, but critics question the tradeoffs
Anthropic has introduced an invisible watermarking system for Claude based on Google's SynthID-Text technology to meet European AI transparency requirements. While this move aims to make AI-generated content detectable, it has prompted criticism regarding potential impacts on the quality of the model's word choices. The development highlights the ongoing struggle to balance regulatory compliance with the technical performance of large language models, creating new challenges for legal teams and researchers verifying the origin of digital text.