Major Publishers Sue Meta Over AI Training

Meta is facing a fresh legal barrage over how it built its AI. Five major publishers—Hachette, Macmillan, McGraw Hill, Elsevier, and Cengage—along with novelist Scott Turow have filed a class-action lawsuit in federal court in Manhattan accusing Meta and CEO Mark Zuckerberg of mass copyright infringement tied to training its Llama AI models. The proposed class action complaint claims Meta used millions of pirated works ranging from novels, textbooks, and scientific articles, scraping them from the web while stripping copyright information, and that Zuckerberg himself approved the strategy, reports the New York Times.

The lawsuit argues Llama can both mimic authors' styles and summarize books so thoroughly that it undercuts sales and allows copycats to flourish. Llama allegedly admitted to being trained on writers' published works, including those of Turow, a best-selling author who says Meta's alleged actions are "shameless, damaging and unjust," per the Times. The publishers want a court order forcing Meta to destroy all unlawfully obtained training data and halt the practices at issue. In response, Meta has vowed to fight the lawsuit "aggressively," noting that "courts have rightly found that training AI on copyrighted material can qualify as fair use," per Reuters. A judge tossed a related copyright infringement lawsuit against Meta last year without absolving the company.