A US judge has ruled that Anthropic can use copyrighted books to train AI, invoking the doctrine of fair use.

A decision by a federal judge in San Francisco allows the firm Anthropic to train its artificial intelligence models with copyrighted books without permission, ruling that this constitutes fair use. The judge compared AI's learning process to humans reading books. However, the decision does not provide a blank check, given that Anthropic also downloaded pirated books, which could constitute copyright infringement. The lawsuit will continue to determine possible damages.