Nearly 300 authors sue Meta over alleged AI book copying

Nearly 300 authors filed five copyright-infringement complaints against Meta Platforms in the U.S. District Court for the Northern District of California on Oct. 19, accusing the company of using pirated and scanned books to train its artificial-intelligence models. The coordinated cases are Winspear, Levine, Reno, Sheard and Older v. Meta Platforms. The authors chose individual claims rather than a class action, though the complaints make substantially the same allegations and were divided largely for logistical reasons. The Sheard complaint alone has 58 authors as plaintiffs. The claims allege that Meta trained its Llama and Muse large language models on copyrighted books obtained from shadow libraries including LibGen, Z-Library and Anna’s Archive. The authors also accuse Meta of using BitTorrent or other torrenting protocols to download works while distributing copies through peer-to-peer networks, and of removing copyright notices and other copyright-management information from files. A second alleged route of copying is Project Vellum, an industrial-scale book-scanning operation that the complaints say began in mid-2025. Meta allegedly acquired physical books, converted them into digital copies with optical character recognition and used the resulting material for AI training without authorization or licences. The lawsuits say Project Vellum supplied training material for Muse Spark and Muse Glimmer. Meta released Muse Spark in April and Muse Glimmer in August. The Authors Guild backed the litigation, describing the alleged mass scanning for AI training as an independent act of infringement. The Sheard complaint alleges direct and contributory copyright infringement under the Copyright Act, as well as violations of the Digital Millennium Copyright Act. The plaintiffs seek statutory and actual damages, profits obtained from the alleged infringement, a permanent injunction, attorneys’ fees and a jury trial. Statutory damages for willful infringement can reach $150,000 per work without proof of actual loss. The cases arrive after Judge Vince Chhabria ruled on June 25 in Kadrey v. Meta that Meta’s use of authors’ books to train AI was fair use, while indicating that a showing of market dilution could have altered the result. The new plaintiffs are expected to put the prospect that AI-generated books reduce sales of human-written works at the centre of their arguments. The authors were recruited through ClaimsHero, which partnered with Freedman Normand Friedland and Stris & Maher to reach writers who opted out of Anthropic’s $1.5 billion books settlement. As we reported in July, that settlement was approved at roughly $3,000 per eligible work. The Meta plaintiffs are seeking substantially higher compensation. Judge William Alsup had criticized ClaimsHero’s effort to persuade plaintiffs to opt out of a settlement at a November 2025 hearing, calling it “a fraud of immense proportions.” Kyle Roche of Freedman Normand Friedland represents the authors with Elizabeth Brannen of Stris & Maher; Roche’s firm represents more than 2,200 authors and planned to file up to 1,000 individual claims by the end of the year. Separate Meta copyright cases brought by Chicken Soup for the Soul, Cognella and Cambronne are scheduled for jury trials beginning May 3, 2027.