A federal magistrate judge has ordered Meta Platforms to identify every person who made, authorised or approved the decisions to obtain pirated book datasets from shadow libraries including LibGen and Anna’s Archive for AI training — rejecting Meta’s argument that the decisions were team consensus with no single decision-maker to name.
The discovery order, dated October 19, was issued by US Magistrate Judge Thomas S. Hixson in Entrepreneur Media’s lawsuit against Meta, one of five jointly managed complaints accusing the company of unlawfully training its AI models on protected writing. It is a disclosure ruling, not a finding of infringement: it decides what Meta must hand over, not whether the company broke copyright law.
What Meta Has To Disclose
The order instructs Meta to identify every “natural person who made, authorized, or approved any decision to obtain or download any Shadow Library Dataset”, together with the circumstances of each decision, each person’s approximate dates of involvement and which decision he or she made. Anna’s Archive and LibGen are named in the order, and LibGen’s SciMag collection — reportedly containing some 80 million scholarly papers and articles — is covered for the torrenting questions.
Meta must also identify the people who gave the green light to obtaining the datasets “by means of the BitTorrent or other peer-to-peer protocol”, including each person who decided whether or how to seed, or to limit seeding, while Meta downloaded the LibGen SciMag and Anna’s Archive content. Seeding matters legally because a torrent downloader who seeds is also distributing the files to other users, which is the basis of the remaining distribution claims in the litigation.
Team Consensus Is Not An Answer, The Judge Ruled
Meta had told the court that it did not necessarily have a single decision-maker, that decisions were often collaborative or the result of team consensus, and that it was unaware of anyone who decided to upload or distribute the datasets. Its earlier answers had named several people believed to be involved in data-related decisions, but said no one “formally” made or approved the decisions to acquire or use books for its first two Llama models.
Judge Hixson ruled that this does not end the inquiry. “If a decision is fairly attributable to a group or a team, everyone in the group or team who participated in making that decision is responsive to this rog,” he wrote, using the shorthand for an interrogatory. The order follows an earlier ruling last month in which the same judge ordered Meta to hand over the command histories of the servers it used to torrent the pirated books — logs that would show how the torrent clients were set up, including their upload settings.
Where The Wider Litigation Stands
The order lands in litigation that has already produced one major ruling for Meta. In June 2025, in the related authors’ case brought by writers including Richard Kadrey and Sarah Silverman, the presiding judge ruled in Meta’s favour on the named plaintiffs’ claim that using their books to train AI models was fair use — while leaving unresolved a claim about alleged distribution during the torrenting itself. That split is why the discovery fight now centres on who approved the downloads and the seeding, rather than on training alone.
The stakes were set by a neighbouring dispute: in Bartz v. Anthropic, the Claude developer agreed to pay $1.5 billion to affected authors over alleged pirated-book training data. Our overview of the copyright fights reshaping AI traces how those cases divide between training and acquisition claims.
Why It Matters
Meta’s torrenting decisions are where its copyright exposure now concentrates. The training question has, for the moment, been answered in the company’s favour in the authors’ case; the acquisition and distribution questions have not. An order that forces the company to attach names, dates and specific decisions to the downloads changes the shape of that fight: individual approvals can be tested against the company’s public position that the choices were diffuse team outcomes, and what Meta discloses in Entrepreneur Media’s case may prove useful to the other plaintiffs in the jointly managed litigation.
It is also the second courtroom story to reach Meta’s AI operation this week, after the California suit over its AI glasses allegedly recording conversations without the wake phrase. How Meta’s teams documented data decisions is becoming a legal question in its own right, alongside the privacy pause after thousands of shared Muse chats were found discoverable in public DNS records.
Background
The shadow-library litigation grew out of revelations that Meta downloaded tens of terabytes of books from Library Genesis, Z-Library and Anna’s Archive through Anna’s Archive, after internal debates about licensing the material instead. Authors sued in 2023 in what became the Kadrey case; publishers and other rightsholders, including Entrepreneur Media, followed with related complaints that are now managed together. Meta has argued that training on publicly available material is fair use, and won that point for the named authors’ books in 2025 — but the torrenting and seeding allegations were never part of that victory, and they are what this discovery order is about.
Source: Digital Music News; TokenPost; TorrentFreak, as republished by GhanaMMA — all reporting Magistrate Judge Thomas S. Hixson’s October 19 discovery order in Entrepreneur Media v. Meta.
0 Comments