Zuckerberg personally approved using millions of books to train Meta’s AI, lawsuit claims

Five major publishers and author Scott Turow have filed a class-action lawsuit against Meta and CEO Mark Zuckerberg, accusing them of illegally copying millions of books and articles to train the company’s Llama AI system. Todd Spangler reports for Variety. The plaintiffs include Hachette, Macmillan, McGraw Hill, Elsevier and Cengage.

The lawsuit, filed in the U.S. District Court for the Southern District of New York, claims Zuckerberg “personally authorized and actively encouraged” the infringement. According to the complaint, Meta torrented over 267 terabytes of pirated material, equivalent to hundreds of millions of publications.

The suit further alleges that Meta initially considered licensing content, with a proposed budget of up to $200 million. However, after the matter was escalated to Zuckerberg in April 2023, the company abandoned that strategy. An internal memo noted that licensing even a single book would undermine a potential fair-use legal defense.

Meta also allegedly stripped copyright management information from the stolen works to conceal their origin.

Meta responded: “Courts have rightly found that training AI on copyrighted material can qualify as fair use. We will fight this lawsuit aggressively.”

This argument has previously succeeded. In June 2025, a federal judge ruled that Meta’s use of nearly 200,000 books to train Llama constituted fair use. The new lawsuit argues that Meta’s deliberate circumvention of copyright protections places its conduct outside fair-use provisions.