A federal judge has ruled that Anthropic’s use of copyrighted books to train its artificial intelligence model constitutes fair use under U.S. copyright law — a significant win for the generative AI industry.
U.S. District Judge William Alsup in San Francisco said the company’s training of its Claude AI model using texts from authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson fell within legal bounds. However, he also ruled that Anthropic’s storage of the books in a centralized database did violate copyright protections.
The class-action lawsuit is one of several challenging the use of copyrighted content by AI developers, including OpenAI, Meta, and Microsoft. These companies argue that using large datasets is necessary to build innovative tools and that fair use covers their practices.
The court’s ruling is the first to address fair use in the context of generative AI, setting an important precedent. Anthropic, backed by Amazon and Alphabet, has not publicly commented on the decision.
While the central library issue may result in limited liability, the broader ruling strengthens the legal foundation for companies using published works to train AI models — a critical component in the race to develop more advanced and responsive language technologies.
