A United States magistrate judge has ordered Meta to identify everyone who took part in the decisions to download shadow library datasets using BitTorrent, after the company told a publisher's lawyers that it could not identify anyone who had formally made or approved them.
The ruling, by Magistrate Judge Thomas Hixson, is the latest in a series of discovery disputes in the lawsuits over Meta's use of shadow libraries. Last month the same judge ordered the company to hand over the command histories of the servers it used to torrent books, logs that would presumably reveal how its torrent clients were set up, including any changes to their upload settings.
In the class action filed by authors including Richard Kadrey and Sarah Silverman, the court has already ruled that training AI models on the books was fair use. That left the BitTorrent claims as the last live part of that case.
The server logs would show what happened on the machines. The publisher Entrepreneur Media, which publishes Entrepreneur magazine and filed one of the related lawsuits, wanted to know something else: who approved the torrenting. Its suit says Meta contributed to piracy by uploading, or seeding, works to third parties.
To find out, the publisher asked Meta to identify everyone who "made, authorized, or approved" the decisions to obtain the shadow library datasets, and to do so by BitTorrent.
Meta's answers named several people believed to have been involved in data-related decisions. The company said, however, that it could not identify anyone who "formally" made or approved decisions to acquire or use books for its first two Llama models, arguing that those decisions were often the result of team consensus. On the torrenting itself it pointed to the employees who ran the downloads and described their choices as technical decisions. It added that the acquisition of data from Anna's Archive went through an internal legal review in early 2024 involving its in-house lawyers.
Entrepreneur Media put the implication plainly in its filing. "On that construction no one at Meta decided to torrent anything, yet Meta torrented shadow libraries repeatedly from 2022-2024," it countered.
The question sharpens around seeding, which is the part of the BitTorrent protocol that sends data back out. An internal Meta logbook, which one employee reviewed before starting a download from Anna's Archive, records a decision not to use torrents. "Decided to go with direct file uploads without using torrents for the following reasons: Using torrents would entail seeding the file such as sharing content outside. This would be legally not okay," the logbook reads, as quoted in the publisher's filing.
Meta does not deny that it used BitTorrent to download LibGen's Scimag collection and data from Anna's Archive. It says it is not aware of anyone who made, authorised or approved a decision to upload or distribute the data, and that measures were implemented to avoid seeding. How much data was shared is unclear: last year the company's own expert estimated that uploads from its AWS servers did not exceed roughly 30 per cent of the data downloaded during torrent downloads between April and July 2024.
Hixson was not persuaded by the company's position that it had nothing more to provide. He noted that the question put to it was not limited to formal decisions, and that its answers were mostly about using the data rather than obtaining or downloading it. Nor did the collaborative nature of the decisions help: the question asks about all decisionmakers, so if a decision was made by a team, everyone involved has to be named.
The order requires Meta to identify the people who took part in the discussion that preceded the 2023 LibGen torrent download, those who approved going ahead with BitTorrent, and whoever gave the go-ahead for the April 2024 download from Anna's Archive. It also covers everyone who decided "whether or how to seed, or to mitigate seeding". Whether the torrenting was fair use, and whether the company actually distributed books to anyone else, has yet to be decided. The dispute was reported by TorrentFreak.
The company's construction has a formal elegance that deserves preserving. The decisions were reached by consensus. A consensus is not reached by an individual. No individual therefore made the decision, and the decision has no author.
The downloads, meanwhile, took place. The publisher's filing compresses the whole matter into a single line: nobody decided to torrent anything, and the torrenting happened repeatedly across three years.
The court's remedy runs the logic back the other way. If a decision made by a team belongs to everyone on the team, then a decision made by nobody belongs to all of them, and the qualification for being a decisionmaker becomes attendance.
The Department of Unnecessary Research draws the practical lesson for the modern workplace. An employee who acts alone has made a decision. An employee who agrees with four colleagues has witnessed one.
On the company's own account, the only participants who can be shown to have done anything are the servers, and they have already surrendered their command history.