AI Training on Copyrighted Works Ruled Fair Use Judge Alsup

A federal judge in San Francisco has ruled that training an AI model on copyrighted works without specific permission does not violate copyright law. U.S. District Judge William Alsup stated that AI company Anthropic could assert a “fair use” defense against copyright claims for training its Claude AI models on copyrighted books. However, the judge emphasized that the method of obtaining these books is crucial.

Judge Alsup ruled that it is not permissible for Anthropic to have downloaded millions of pirated titles and maintain a digital library of them. The judge ordered a separate trial to address Anthropic’s storage of these pirated books and has not yet decided whether to grant the case class action status.

This ruling addresses a long-standing question in the AI industry: Can copyrighted data be used to train generative AI models without the owner’s consent? Since the launch of OpenAI’s ChatGPT in 2022, dozens of lawsuits have been filed questioning the legal “fair use” of content in AI training. Alsup’s decision may set a precedent for these cases, although many rulings are likely to be appealed, potentially delaying clarity on AI and copyright law in the U.S. for years.