US judge orders Meta to name employees behind torrenting pirated books for AI

Meta has to officially name the employees who greenlit the use of BitTorrent to download pirated books for training its AI models. US Magistrate Judge Thomas Hixson has ordered the tech giant to identify everyone involved in sourcing data from shadow libraries like LibGen and Anna's Archive.This order goes beyond the engineers who simply ran the downloads. Meta must reveal the names of all team members who decided to use torrents and whether to "seed" (upload) those files to others. To figure out what happened on Meta's machines, the judge previously ordered the company to hand over the command histories of its servers. These logs will reveal exactly how the torrent software was set up and if any upload settings were tweaked.The fair use rulingThis dispute is part of a larger copyright lawsuit filed by publishers like Entrepreneur Media and authors such as Sarah Silverman and Richard Kadrey. Interestingly, the court has already ruled that training AI on these books falls under 'fair use'. That leaves the BitTorrent claims as the last major part of the case.When you download a file via BitTorrent, the software automatically uploads parts of that file to other users on the network. Because of this, publishers accuse Meta of actively distributing their copyrighted work. Meta's own expert estimated that up to 30% of the data downloaded between April and July 2024 was uploaded from its AWS servers. "Legally not okay"When publishers asked who authorised the torrenting, Meta tried to dodge the question. The company argued that decisions for its first two Llama models were based on "team consensus" and described the downloads as mere "technical decisions" made by engineers. Meta also noted that an internal legal review involving in-house lawyers took place in early 2024, but it did not name any higher-ups who approved the torrenting.However, court filings revealed an internal logbook where an employee discussed an upcoming download from Anna's Archive. The employee explicitly noted a decision to avoid torrents because seeding the files outside the company would be "legally not okay." Despite this clear internal warning, Meta repeatedly torrented data from these shadow libraries between 2022 and 2024. Judge Hixson was not convinced by Meta's "team consensus" defence. He ruled that if a decision was collaborative, every single person in that group must be named. Meta now has to identify everyone who participated in discussions ahead of the 2023 LibGen download and whoever approved the April 2024 download from Anna's Archive.The massive scale of downloadsThe scale of the pirated data is massive, as the decision-making wasn't just limited to lower-level staff. The involvement reportedly reached director and vice president levels at Meta.This ongoing legal battle highlights a growing trend of Big Tech companies facing massive consequences for how they scrape data. As publishers continue to fight for their rights, Meta’s unmasking of its employees could have major ripple effects across the entire AI industry.- Ends