Five leading publishers are filing a lawsuit against Meta regarding Llama.

      Elsevier, Cengage, Hachette, Macmillan, and McGraw Hill, along with author Scott Turow, filed a proposed class action lawsuit in Manhattan on Tuesday, claiming that Meta unlawfully used millions of their works to train Llama. Following Judge Chhabria's ruling in June 2025, plaintiffs who possess stronger evidence of market harm have been anticipating their opportunity.

      On Tuesday morning, five of the largest publishers in the world, alongside a prominent American novelist, entered a federal courthouse in Manhattan to file a proposed class action complaint against Meta Platforms. According to Reuters, the complaint alleges that Meta illegally used millions of their books and journal articles to train its Llama large language models without obtaining permission, payment, or a license. The lawsuit requests the court to classify it as a class action representing all rights holders in similar circumstances.

      This filing is the most recent addition to a series of copyright cases concerning AI training. However, it differs significantly from previous cases in substance.

      Why This Case Differs from Kadrey

      Those familiar with AI-related copyright litigation will recognize the precedent case: Kadrey v. Meta. Filed in 2023 in the Northern District of California by authors including Sarah Silverman, Richard Kadrey, Christopher Golden, Ta-Nehisi Coates, Junot Díaz, and Michael Chabon, that case made similar allegations—that Meta retrieved copyrighted works from piracy sites (LibGen, Z-Library, and Anna’s Archive) and used them for training Llama.

      Court documents referenced by Tom’s Hardware revealed that Meta employees downloaded approximately 82 terabytes of pirated content during this process. Mark Zuckerberg personally approved the use of LibGen for Llama's training, despite internal AI leaders indicating that it consisted of pirated data, which could weaken Meta’s discussions with regulators.

      Ultimately, Meta won that case. In June 2025, Judge Vince Chhabria ruled in favor of Meta on fair-use grounds, concluding that training Llama with copyrighted books was transformative enough to meet the fair-use standard. However, Chhabria's ruling was unusually narrow and candid regarding its limitations.

      He publicly stated that Meta's victory "may be in significant tension with reality" and that the ruling applied solely to the specific authors involved in the case. He expressly pointed out that future plaintiffs could prevail if they provided stronger evidence of market harm, which, in his opinion, the Kadrey plaintiffs lacked.

      The filing on Tuesday appears, at first glance, to align with the type of case Chhabria suggested.

      What the Publishers Are Presenting That the Authors Did Not

      There are three structural distinctions between Kadrey and this new lawsuit, all of which favor the plaintiffs. The first difference is the scope of the catalogue. Kadrey involved around 666 specific titles from a limited number of individual authors, while the new complaint encompasses the complete publishing operations of five companies that collectively represent a significant portion of global academic, educational, and trade publishing.

      According to Reuters's description, the titles include not only literary works like N.K. Jemisin's "The Fifth Season" and Peter Brown's "The Wild Robot" but also textbooks, scientific journal articles, and reference materials. The market dynamics for these works, especially in the academic and educational sectors, differ fundamentally from the trade-fiction market dominated by the Kadrey plaintiffs.

      The second difference lies in the evidence of market harm. Academic and educational publishers can specify documented revenue streams that AI-trained models replace in ways individual authors typically cannot. When Llama provides an answer to a student's biology question that would have otherwise required consulting a Cengage textbook, the substitution is straightforward and quantifiable.

      The plaintiffs are likely to present this substitution as a clear instance of market harm that Chhabria's earlier ruling indicated was missing from Kadrey. Reed Smith's analysis of recent fair-use judgements noted that the market harm aspect, more than transformative use, is becoming the focal legal battleground.

      The third difference pertains to the licensing context in the market. Since 2023, AI companies have increasingly entered licensing agreements with publishers. Meta itself has signed such deals with Reuters, CNN, Fox News, People Inc., and USA Today.

      The existence of these licenses is a significant factor in fair-use law: courts evaluating market harm now have evidence that a licensing market exists, with some publishers negotiating and pricing their participation. The new plaintiffs will argue that Meta's decision to bypass negotiations with them while opting to license others demonstrates bad faith.

      The Anthropic Settlement in the Background

      The case on Tuesday occurs in light of another recent precedent. Earlier this year, Anthropic agreed to compensate authors as part of resolving the Bartz v. Anthropic class action over similar claims, which the Authors Guild characterized as a significant settlement. The details and amount set a precedent for what AI-training copyright cases could yield when they reach financial resolution rather than early summary judgment.

      TNW has monitored Anthropic’s broader business trajectory, including its $1.