Anthropic illegally trained Claude with Stanford professors' books (stanforddaily.com)

Anthropic, the AI company behind the Claude models, has faced a major legal setback after a federal judge ruled that the company illegally trained its models using copyrighted texts from over 10 Stanford professors. In the May 2026 lawsuit Bartz v. Anthropic, Judge William Alsup found that Anthropic had pirated works from authors including Tobias Wolff and Abraham Verghese by downloading and storing their texts from unauthorized sources. This ruling is significant for the AI/ML community as it establishes a precedent regarding copyright issues in AI training, highlighting the legal complexities of utilizing copyrighted materials without permission. The court's decision also limits Anthropic’s ability to create a central library of texts for AI training. The $1.5 billion settlement awarded to authors and publishers affects nearly 500,000 registered works, but some authors, like Wolff and Chang-rae Lee, have opted out, seeking greater compensation and legal recourse in a separate lawsuit. They argue that current settlements do not adequately protect their intellectual property rights or ensure fair use in AI training practices. The ongoing legal battles reflect broader concerns among authors and copyright advocates about the implications of AI technologies on creative work, including potential risks to the livelihoods of future authors. As litigation continues, there is a push for clearer legislation requiring AI companies to license copyrighted materials, which could reshape the future landscape of AI model training.