Particle.news

Download on the App Store

Judge Rules AI Training on Copyrighted Books as Fair Use, Faults Pirated Library

U.S. District Judge William Alsup’s ruling spares Anthropic’s AI training on purchased books from liability; it sends the company’s storage of more than seven million pirated copies to a December trial.

Image
Image
Anthropic wins AI copyright ruling, judge says training on purchased books is fair use

Overview

  • Alsup found that copying and digitizing legally purchased books to train Anthropic’s Claude model was “exceedingly transformative” and qualifies as fair use under Section 107 of the Copyright Act.
  • The court held that Anthropic’s retention of over seven million pirated book copies in a permanent digital library infringes copyright and lies outside fair use protections.
  • A jury trial is set for December 2025 to assess damages for the pirated-library infringement, with potential penalties of up to $150,000 per willfully infringed work.
  • Legal experts say this is the first comprehensive fair use ruling in a generative AI case and that it will influence dozens of pending lawsuits against companies like OpenAI and Meta.
  • The decision highlights the importance for AI developers to source training materials lawfully to avoid future infringement claims.