Is it legal to train AI models on copyrighted books? It’s complicated

The U.S. District Court ruled that Anthropic’s training of its language model was lawful, but imposed a $1.5 billion penalty because the company obtained the books from illegal shadow libraries rather than through authorized channels. Attorneys note that the key legal question is whether AI training constitutes “fair use” or a direct copy that competes with the original market, a determination still unsettled by copyright law that has not been updated since 1976.