tech

Anthropic rolled out a feature that stops undetected AI-generated writing

Anthropic said new Claude models will embed an "imperceptible watermark" directly into AI-generated text.

Anthropic rolled out a feature that stops undetected AI-generated writing

TL;DR

  • Anthropic's new Claude models will embed an imperceptible watermark into AI-generated text.
  • The watermark is designed to be undetectable to humans but traceable by tools.
  • This feature supports transparency commitments under the European Union AI Act.
  • The move could impact the publishing industry, which has faced issues with AI-generated content in novels.
  • Heavy editing, paraphrasing, or translation can make the watermark undetectable.
  • Google DeepMind has also implemented watermarking for its Gemini AI outputs.