The AI copyright fight lost on "training." So the courts have moved on to "who clicked download."
Updated:
A federal magistrate judge has ordered Meta to identify, by name, every employee who made or approved the decision to obtain copyrighted works through shadow libraries including Anna's Archive and LibGen. The order, finalized October 19 by Magistrate Judge Thomas S. Hixson in the sprawling Meta books litigation, instructs the company to produce each "natural person who made, authorized, or approved any decision to obtain or download any Shadow Library Dataset" (Digital Music News).
Meta's hoped-for dodge was diffusion of responsibility — the company argued it "did not necessarily have a single decisionmaker and that decisions were often collaborative or the result of team consensus." Judge Hixson's answer turns that into an instruction: "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" (Digital Music News). The order also reaches the names behind BitTorrent use — specifically, "each person who made, authorized or approved the decision whether or how to seed, or to mitigate seeding, when Meta obtained LibGen Scimag [reportedly containing some 80 million scholarly papers and articles] and Anna's Archive content by BitTorrent."
Why this is where the litigation actually lives
The discovery order is technically part of one of five jointly managed complaints accusing Meta of training AI models on protected writing. But the substantive action has moved to the acquisition pipeline, because the training question has gone badly for plaintiffs. In June 2025, two California federal judges — in the Anthropic and Meta book cases — handed AI companies fair-use wins on the core question of ingesting unlicensed works for training, rulings that "sent shockwaves" among stakeholders as the first fair-use decisions of the AI era (Billboard).
Yet those same rulings left the door open on a different theory: both were "clear in leaving room for claims about pirated acquisition," and the cases moved toward discovery on how the datasets were actually obtained (Billboard). That is the seam this order now pulls. And last July, the pressure showed its first major settlement: Anthropic agreed to pay $1.5 billion to authors whose works its Claude models had allegedly been trained on — a piracy-driven settlement, not a training-loss verdict (Digital Music News).
The music business is reading the docket closely
Music rightsholders are watching the books cases as a preview of their own playbook. When Sony Music Publishing and Warner Chappell sued Anthropic in August 2026, they claimed a "brazen campaign of illegally torrenting, scraping, and downloading copyrighted works on a massive scale" and named not just the company but founder Dario Amodei and co-founder Benjamin Mann as defendants (Billboard). Anthropic's response — "training generative AI models is a transformative fair use," calling the suit "recycling allegations" (Billboard) — is precisely the defense that succeeded on training and failed nowhere yet on acquisition.
The individual-employees theory is the music industry's leverage point. If discovery in a books case can establish which humans green-lit the torrenting of shadow libraries, those names, dates and internal decisions become usable by every other plaintiff — and Meta's own music docket is already crowded: a long-running fight with Epidemic Sound, plus copyright clashes with Eminem's publisher Eight Mile Style and Wixen Music Publishing (Digital Music News).
The counterweight
There is a real argument on the other side. Fair use on training means the publishers' whole case now rests on sourcing, and if courts ultimately treat shadow-library acquisition as the industry-standard practice it arguably was — datasets that everyone in the field downloaded — the individualized finger-pointing could look like process abuse to an appellate panel. The June 2025 rulings are already headed upward: district decisions "are not precedent," and more than 40 pending AI copyright suits will reach circuit courts and eventually the Supreme Court (Billboard). Until then, the name-gathering goes forward — and for Meta's lawyers, the uncomfortable part of this week's order is that the list can only grow with each "collaborative decision" the company describes.
Sources
- Digital Music News — Federal Judge Orders Meta to Identify the Employees Who Authorized Shadow Library Piracy (Dylan Smith), October 20, 2026
- Billboard — Sony & Warner Sue Anthropic in Latest AI Music Copyright Lawsuit: 'Illegally Harvested to Fuel Claude' (Bill Donahue), August 31, 2026
- Billboard — AI Legal Analysis: What Book Copyright Rulings Mean for Music Lawsuits (Rachel Scharf), July 14, 2025