Meta tries to scapegoat staff who downloaded torrents to spare execs

Litigation over use of copyrighted books to train Meta’s AI centers on allegations that Meta downloaded massive troves from shadow libraries like LibGen and Anna’s Archive and that responsibility runs above rank-and-file engineers. Plaintiffs - authors including Richard Kadrey and Sarah Silverman and publisher Entrepreneur Media - cite testimony that Meta pulled 80.6 TB from LibGen and 35.7 TB from Anna’s Archive, internal messages showing staff discomfort and requests for director/VP approval, and at least one unredacted string referencing an escalation to “MZ” (Mark Zuckerberg). Magistrate Judge Thomas Hixon has ordered Meta to identify every natural person who “made, authorized, or approved” any decision to obtain or download those shadow-library datasets and to produce command history files and server logs (including virtual machines and AWS instances) used for torrenting, which will let plaintiffs prove whether Meta seeded specific works and how its claimed seeding mitigations operated.

Meta defended itself by arguing early dataset compilation was “informal,” that no formal executive decision authorized piracy, and that plaintiffs’ discovery requests were overbroad; it also denied awareness of anyone who approved uploads or distributions. Hixon rejected that narrow construction, required detailed identifications (names, roles, dates, and what each person authorized or contributed), and compelled answers about pirated works not used for training. The order tightens discovery and shifts scrutiny up the chain of command, potentially exposing supervisors, managers, and executives who greenlit or tacitly approved torrenting for AI training.