Magistrate Judge Thomas Hixon has ordered Meta to identify the executives and managers who authorized the use of torrents for AI training. This ruling stems from class-action lawsuits filed by authors, including Richard Kadrey and Sarah Silverman, and publisher Entrepreneur Media, who allege that Meta illegally used copyrighted works.
Meta previously claimed that its early AI training datasets were compiled through an informal process without formal leadership approval. The company argued that no specific authority directed the downloading of 80.6 terabytes of data from LibGen or 35.7 terabytes from Anna’s Archive between 2022 and 2024.
Internal evidence suggests otherwise, as some staff members testified that downloads required director or VP-level approval. One unredacted message indicated that the decision to download from LibGen was escalated to Mark Zuckerberg. Meta has consistently denied that any formal decision-making process for torrenting existed.
The court order also mandates that Meta produce command history files for all servers, virtual machines, and AWS instances used during the process. These logs are expected to clarify whether the company engaged in seeding copyrighted content, a claim Meta denies by citing the use of mitigation measures.
Meta must now provide the names of individuals involved, their roles, and the dates of their involvement regarding the acquisition of shadow library datasets. The judge rejected Meta's attempt to provide only a list of names without context, as well as its argument that decision-making was too collaborative to identify specific individuals.
