A federal magistrate judge has ordered Meta to disclose the names of employees who authorized, approved, or participated in decisions to download copyrighted books from shadow libraries using BitTorrent for its Llama artificial intelligence models.
The ruling marks the latest development in an ongoing legal battle concerning copyright infringement and data acquisition practices in the tech industry. While courts have previously ruled that training AI models on copyrighted books falls under fair use in certain contexts, Meta’s reliance on BitTorrent to acquire these datasets remains a live and contentious issue in multiple class-action and individual lawsuits.
Last month, Magistrate Judge Thomas Hixson ordered Meta to hand over the command histories of the servers it used to torrent pirated books. These server logs are expected to reveal how Meta’s torrent clients were configured, including any modifications made to upload and sharing settings. However, publishers seeking accountability argue that knowing how the servers operated is only part of the puzzle. They also want to know exactly who inside the corporate hierarchy greenlit the activity.
The dispute over corporate accountability has brought intense scrutiny to how large technology companies source data for foundational models. As publishers and authors press forward with contributory infringement and distribution claims, the question of individual and collective decision-making within Meta has taken center stage.
‘Technical’ Decisions and Corporate Consensus
The current discovery dispute involves a lawsuit filed by Entrepreneur Media, the publisher of Entrepreneur magazine. The publisher accuses Meta of contributing to copyright infringement by uploading, or seeding, protected works to third parties via BitTorrent networks while downloading datasets for its AI development. This mirrors similar contributory infringement claims brought by authors such as Richard Kadrey and Sarah Silverman in a separate class-action lawsuit.
To investigate the chain of command, Entrepreneur Media asked Meta to identify every person who made, authorized, or approved decisions to obtain shadow library datasets using BitTorrent protocols. In response, Meta named several employees who were involved in data-related operations. However, the tech giant drew a distinct line regarding formal approval.
Meta argued that it could not identify anyone who formally made or approved decisions to acquire or use books for its first two Llama models. According to the company, those decisions were typically the result of a collective team consensus rather than a top-down directive from a single executive. When pressed specifically on the torrenting activity, Meta pointed to the engineers who executed the downloads, describing their actions as routine technical decisions rather than policy-level approvals.
Furthermore, Meta noted that the acquisition of data from platforms like Anna’s Archive underwent an internal legal review in early 2024, involving members of its in-house legal team. By framing the actions as decentralized team choices and technical executions, Meta avoided naming higher-ranking executives who might have sanctioned the use of peer-to-peer file-sharing networks.
This defense drew sharp criticism from the plaintiffs. Entrepreneur Media countered that under Meta’s construction of events, no one at the company decided to torrent anything, yet the company systematically torrented shadow libraries repeatedly between 2022 and 2024.
‘Legally Not Okay’
The friction between Meta’s official narrative and internal corporate documentation forms a central pillar of the plaintiffs’ arguments. Evidence uncovered during the discovery process suggests that employees were well aware of the legal risks associated with peer-to-peer file sharing and BitTorrent protocols.
Testimony from an engineer who participated in downloading data from Library Genesis (LibGen) revealed that the issue of seeding was raised and debated internally. Crucially, the engineer noted that discussions included steps intended to limit outbound sharing.
An internal Meta logkeeper reviewed before initiating a download from Anna’s Archive explicitly flagged the dangers of using torrent networks. According to excerpts cited in Entrepreneur Media’s legal filings, the logbook entry stated that the team decided to bypass torrents due to the inherent mechanics of the protocol.
The logbook reads that using torrents would entail seeding the file, thereby sharing content externally with third parties on the network. The entry concluded that this outcome would be "legally not okay."
Meta does not dispute that it utilized BitTorrent to download the LibGen Scimag collection and datasets from Anna’s Archive. However, the company maintains that it is unaware of any individual who explicitly made, authorized, or approved a directive to upload or distribute that data to the wider public. Meta’s legal representatives have emphasized that various measures were implemented specifically to avoid seeding.
The exact volume of data shared with external parties remains a point of contention. Last year, an expert retained by Meta estimated that uploads from the company’s Amazon Web Services (AWS) servers accounted for roughly 30 percent of the total data volume downloaded during torrent operations conducted between April and July 2024.
Everyone in the Group
The impasse reached a head when Magistrate Judge Hixson issued his ruling on the dispute. Two weeks prior to the decision, Meta informed the court that it possessed no additional information to provide regarding the internal decision-making process behind the torrenting activities.
Judge Hixson rejected Meta’s stance, pointing out that the interrogatories submitted by the plaintiffs were not restricted solely to formal, top-down decisions. The judge observed that Meta’s previous answers focused heavily on the subsequent use of the data rather than its initial acquisition and downloading.
Addressing Meta’s argument that decisions were frequently collaborative, the judge ruled that collaborative processes do not exempt a company from identifying decision-makers. The interrogatory explicitly asked for all individuals involved in the decision, meaning that if a choice was made by a team, every member of that group must be named.
Under the court order, Meta is required to identify the individuals who participated in the internal discussions preceding the 2023 LibGen torrent download. The company must also name those who approved the transition to BitTorrent and whoever greenlit the subsequent download from Anna’s Archive in April 2024.
Significantly, the mandate extends to anyone who decided whether or how to seed data, or how to mitigate seeding, during the acquisition of the LibGen Scimag collection and Anna’s Archive.
For Entrepreneur Media and other plaintiffs, securing these names is crucial. The publisher argues that the identities of the participants bear directly on the question of willfulness, as well as supporting broader claims of contributory copyright infringement and unauthorized distribution.
While this latest ruling resolves the immediate dispute over naming individuals involved in the data acquisition chain, broader questions remain. The core legal determinations—including whether Meta’s torrenting practices constitute fair use and whether the company illegally distributed pirated books to external parties—have yet to be decided by the court.
Leave a Reply