IP Law Daily, COPYRIGHT NEWS: Publishers, author Turow sue Meta over alleged piracy of books for AI training, (May 6, 2026)

By Saurabh Kashyap, B.A., M.A., LL.B., LL.M.

The complaint alleges that Meta torrented millions of copyrighted books and journal articles from pirate libraries and used them to train its Llama AI models without authorization or compensation.

Major publishing companies and bestselling author Scott Turow have filed a proposed class-action copyright lawsuit against Meta Platforms, Inc., and its founder and CEO, Mark Zuckerberg, alleging that the company unlawfully copied and distributed millions of copyrighted books and journal articles to develop and train Meta’s Llama artificial intelligence models. The complaint alleges violations of the Copyright Act and the Digital Millennium Copyright Act arising from Meta’s alleged torrenting of copyrighted works from pirate databases, downloading of web-scraped datasets, unauthorized reproduction of literary works during AI training, and removal of copyright management information. The plaintiffs seek injunctive relief, statutory and actual damages, disgorgement of profits, destruction of infringing copies, attorneys’ fees, and class certification on behalf of copyright owners whose works were allegedly used without authorization (Elsevier Inc. v. Meta Platforms, Inc., No. 1:26-cv-03689 (S.D.N.Y. May 5, 2026)).

The plaintiffs include Elsevier Inc., Cengage Learning, Inc., Hachette Book Group, Inc., Macmillan Publishing Group, LLC d/b/a Macmillan Publishers, McGraw-Hill LLC, bestselling legal thriller author Scott Turow, and S.C.R.I.B.E., Inc. According to the complaint, the plaintiffs own or control copyrights in millions of literary works spanning fiction, nonfiction, educational textbooks, and scholarly journal articles. The complaint alleges that Meta and Zuckerberg reproduced and distributed those works without authorization to develop Meta’s Llama generative AI models and related AI-powered products.

The complaint alleges that Meta pursued an aggressive strategy to obtain high-quality written materials for training Llama and intentionally bypassed established licensing markets for books and journals. According to the plaintiffs, Meta initially considered licensing copyrighted works from publishers but later abandoned that approach after internal escalations to Zuckerberg. The complaint alleges that Meta instead obtained pirated works from notorious online repositories, including LibGen, Anna’s Archive, Z-Library, Books3, Sci-Hub, and other torrent-based collections. Plaintiffs allege that Meta employees internally acknowledged that the datasets contained pirated works and expressed concerns about their legality and ethics.

According to the complaint, Meta allegedly downloaded and distributed copyrighted works through BitTorrent-based peer-to-peer networks. Plaintiffs contend that Meta torrented over two million copyrighted publications from LibGen in 2022 and later continued downloading additional pirate datasets in 2023 and 2024. The filing alleges that Meta ultimately downloaded more than 267 terabytes of pirated material, which the plaintiffs equate to hundreds of millions of publications. The complaint further alleges that Meta failed to disable BitTorrent’s default file-sharing settings, thereby redistributing copyrighted materials to other users while downloading them.

The plaintiffs also allege that Meta copied copyrighted works from large web-scraped datasets, including Common Crawl, CCNet, and C4. According to the complaint, those datasets contained copyrighted books, journal articles, and subscription-only content scraped from websites and online libraries. The complaint alleges that Meta curated and filtered those datasets for use in AI training while knowingly retaining copyrighted content.

The lawsuit further alleges that Meta repeatedly reproduced copyrighted works throughout the AI training process itself. According to the complaint, preparing literary works for Llama’s training required repeated copying into memory, tokenization, processing pipelines, and training datasets. Plaintiffs contend that each stage of the process constituted a separate unauthorized reproduction under the Copyright Act. The complaint also alleges that Llama memorized portions of copyrighted works and could generate verbatim or near-verbatim excerpts, summaries, derivative works, sequels, study guides, and imitations of copyrighted books and articles.

The complaint includes examples of alleged outputs generated by Llama using copyrighted works. Plaintiffs allege that Llama reproduced passages from Cengage’s Calculus: Early Transcendentals by James Stewart in near-verbatim form. The complaint also alleges that Llama generated summaries and imitations of works by authors including Scott Turow, Sylvia Day, V.E. Schwab, and Becky Lomax. According to the plaintiffs, those outputs function as substitutes for the original works and threaten existing and emerging markets for AI training data licensing.

In addition to copyright infringement claims, the plaintiffs accuse Meta of violating the Digital Millennium Copyright Act by allegedly removing copyright management information from copyrighted works. According to the complaint, Meta intentionally stripped author names, copyright notices, and publication information from datasets obtained from LibGen and Books3 to conceal the use of pirated materials and make it more difficult for copyright owners to identify infringed works. The complaint alleges that Meta retained such information for public-domain works while removing it only from copyrighted materials, demonstrating deliberate conduct.

The filing identifies numerous representative works allegedly infringed by Meta, including textbooks, scientific journal articles, novels, and nonfiction books. Among the cited works are Presumed Innocent, Innocent, and Testimony by Scott Turow; The Fifth Season by N.K. Jemisin; The Wild Robot by Peter Brown; A Darker Shade of Magic by V.E. Schwab; and textbooks published by Elsevier, Cengage, and McGraw-Hill. Plaintiffs contend that these works constitute only a small subset of the millions of works Meta allegedly copied.

The complaint alleges that Zuckerberg personally directed and approved Meta’s acquisition and use of pirated datasets. Plaintiffs claim that Meta employees escalated the question of whether the company should license or pirate copyrighted works to Zuckerberg, after which licensing discussions ceased and Meta proceeded with torrenting activities. The lawsuit asserts claims against Zuckerberg for contributory copyright infringement based on his alleged authorization and encouragement of Meta’s conduct.

According to the complaint, Meta has earned substantial revenue from AI products incorporating Llama and expects AI-related revenue to grow dramatically over the next decade. Plaintiffs allege that Meta integrated Llama into Facebook, Instagram, Messenger, WhatsApp, Meta AI applications, and other products, and that the company’s AI-driven business growth was built upon unauthorized use of copyrighted literary works.

The plaintiffs seek certification of a class consisting of owners of registered copyrights in books and journal articles allegedly reproduced or distributed by Meta through torrenting, web scraping, or AI training. They further seek statutory damages, actual damages and profits, permanent injunctive relief, disclosure of Meta’s training materials and methods, destruction of infringing copies, and attorneys’ fees and costs.

The Case is No. 1:26-cv-03689.

Judge: Castel, P.

Attorneys: Jeffrey M. Gould (Oppenheim & Zebrak, LLP) for Elsevier Inc., Cengage Learning, Inc. and Hachette Book Group, Inc.

Companies: Elsevier Inc., Cengage Learning, Inc., and Hachette Book Group, Inc.; Meta Platforms, Inc.

News: Copyright AINews TechnologyInternet GCNNews