Judge rules Anthropic illegally trained Claude on pirated books

• Judge William Alsup ruled in Bartz v. Anthropic that Anthropic illegally scraped and pirated copyrighted works from over 10 Stanford professors to train its Claude models.
• The ruling prohibits Anthropic from pirating books to create a permanent central library, while justifying the legal acquisition of books for LLM training as fair use.
• The $1.5 billion settlement covers nearly 500,000 registered works, but authors like Tobias Wolff and Chang-Rae Lee opted out to seek up to $150,000 in statutory damages per infringed work in the separate Cruz v. Anthropic lawsuit.
• The Authors Guild is pursuing additional lawsuits against OpenAI and Microsoft and working with Congress to create national legislation requiring licensing for AI training data.

Legal Ruling Details

• Judge Alsup compared AI training to human reading, stating LLMs train on works to create something different rather than to replicate or supplant them.
• The Bartz settlement is described by the Authors Alliance as the largest copyright settlement involving an AI company.

Author Response and Future Litigation

• Tobias Wolff, a professor emeritus of English, stated that using his work without payment is a form of slave labor and that the settlement gives Anthropic too much freedom to use literature.
• Plaintiffs in Cruz v. Anthropic allege that Anthropic illegally downloaded and stored pirated files from shadow libraries like LibGen and PiLiMi in a central library.
• If successful, the Cruz plaintiffs would gain a permanent injunction against Claude model training using their works.