Major Publishers and Author Sue Meta Over Use of Copyrighted Works to Train Llama AI

Five major publishing houses—Elsevier, Cengage, Hachette, Macmillan, and McGraw Hill—along with author Scott Turow filed a lawsuit against Meta on Tuesday in federal court in Manhattan. The proposed class-action complaint alleges that Meta pirated millions of copyrighted works, ranging from textbooks to scientific articles to novels, to train its large language model Llama.

Plaintiffs claim Meta CEO Mark Zuckerberg personally authorized and actively encouraged the copyright infringement.

The lawsuit accuses Meta of ripping copyrighted material from pirate sites including LibGen, Anna’s Archive, Sci-Hub, and Sci-Mag. It further alleges that Meta trained Llama using information from the Common Crawl dataset, which contains unauthorized copies of copyrighted works. As a result, Llama outputs verbatim and near-verbatim substitutes of copyrighted material, according to the complaint.

For instance, when prompted with two brief sentences from Cengage’s textbook Calculus: Early Transcendentals, 9th edition by James Stewart, Llama reproduced word-for-word the continuation of the section. K. Jemisin and The Wild Robot by Peter Brown.

Authors published by the five suing companies include Scott Turow, James Patterson, Donna Tartt, former President Joe Biden, Yiyun Li, and Amanda Vaill. Yiyun Li and Amanda Vaill are two of the Pulitzer Prize winners announced on Monday. The plaintiffs asked the court for permission to represent a larger class of copyright owners.

They seek an unspecified amount of monetary damages and ask the court to order Meta to block its allegedly unlawful activities. The plaintiffs also ask the court to require Meta to provide a list of books, journal articles, and other copyrighted works used to train Llama AI models. Meta stated it will fight the lawsuit aggressively.