Macmillan, Hachette, Elsevier Sue Meta Over Alleged Pirated Content Use in Llama AI Training

Five of the largest publishing houses in the United States have filed a class action lawsuit against Meta Platforms, alleging the company systematically used pirated books and journal articles to train its Llama artificial intelligence model. The complaint, lodged in the U.S. District Court for the Southern District of New York on May 5, 2026, names Elsevier, Cengage, Hachette, Macmillan, and McGraw-Hill as plaintiffs, alongside bestselling author Scott Turow and his company S.C.R.I.B.E., Inc. The publishers accuse Meta of downloading copyrighted material without authorization, scraping web content illegally, and repeatedly leveraging these resources across successive versions of Llama—all without paying any licensing fees.

The scope of the alleged infringement encompasses textbooks, scientific journals, and popular fiction titles. According to the complaint, infringed works include N.K. Jemisin's Hugo Award-winning novel "The Fifth Season" and Peter Brown's children's bestseller "The Wild Robot." A Meta spokesperson rejected the allegations, pointing to prior court rulings that have supported AI training as fair use. The company has not yet formally responded to the complaint in court.

The lawsuit represents a significant escalation in the publishing industry's campaign to hold AI companies accountable for training data practices. Publishers argue that using copyrighted works to train commercial AI models without compensation undermines their ability to monetize intellectual property in an increasingly AI-driven market. Legal observers note the case could set a pivotal precedent for how AI companies source and pay for training materials, potentially forcing fundamental changes to industry practices if the plaintiffs prevail.