Meta has spent weeks pointing at rank-and-file staff over the torrenting of pirated books used to train its AI models. A federal judge now wants the names of the people above them.
Magistrate Judge Thomas Hixson has ordered Meta Platforms to identify everyone involved in deciding to use BitTorrent to pull datasets from shadow libraries like Anna’s Archive and Library Genesis, better known as LibGen. The order, first reported by TorrentFreak, could move the spotlight from the engineers who ran the downloads to the managers and executives who allegedly signed off.
What the court ordered
The ruling, dated October 20, 2026, is a discovery order. Meta must clearly identify the key decision makers behind the choice to torrent books from Anna’s Archive and LibGen.
The order also reaches discussions about seeding during downloads, particularly involving LibGen’s Scimag collection. Seeding means a torrent user uploads pieces of a file to other users while downloading it. In plain terms, a seeder is not just taking copies but also handing them out.
The order follows testimony from Meta employees who handled the torrenting. Those staffers said supervisors, managers, and executives greenlit the activity, which is exactly the chain of command the plaintiffs now want on the record.
Meta’s defense, and why it is wobbling
Meta’s position so far has leaned on two arguments. The company said the torrenting was carried out by lower-level employees, and that many decisions came out of team consensus rather than formal approval from higher management.
Meta previously disclosed some data-related staff involved in the work. It also claimed no formal approvals for the torrenting were documented, even though internal reservations about the legality of the practice existed.
This ruling also builds on earlier discovery fights. In September 2026, Meta was directed to provide server command histories for the machines used in the torrenting.
The scale of the downloads
Meta allegedly downloaded over 81 TB from Anna’s Archive alone, along with tens of additional terabytes from LibGen. The downloads reportedly took place between 2022 and 2024.
The lawsuits behind the order
The discovery order affects two cases in the Northern District of California. The first is a class action known as Kadrey v. Meta, brought by book authors including Richard Kadrey and comedian Sarah Silverman. The second is a separate lawsuit from publisher Entrepreneur Media. Both target the way Meta gathered material to train its Llama large language models.
Early court decisions have concluded that training AI models such as Llama on these materials qualifies as fair use. The remaining dispute centers on the act of downloading the files in the first place, and on any seeding that occurred along the way.
The plaintiffs argue those acts amount to willful infringement and contributory copyright infringement. Willful infringement means the defendant knew it was violating copyright and did it anyway. Contributory infringement means helping someone else infringe, which is where seeding comes in.
What this means
For Meta, the stakes have changed shape. If AI training itself is protected as fair use, the company’s exposure rests heavily on how it acquired the books and whether it redistributed them.
Willfulness is about knowledge and intent, and intent lives in the heads of specific people. A list of names, combined with internal reservations about legality, gives the plaintiffs a far more concrete case to argue that Meta acted knowingly.
The seeding question, especially around LibGen’s Scimag collection, may prove to be the most consequential thread in the case.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.