The court wants the decision chain identified
Magistrate Judge Thomas Hixon ruled on discovery requests in a class action brought by book authors, including Richard Kadrey and Sarah Silverman, and a separate lawsuit from publisher Entrepreneur Media. The order requires Meta to answer questions about people who “made, authorized, or approved any decision to obtain or download the shadow libraries.”
The authors and publisher had asked Meta to identify each person involved, explain their role and dates of involvement, and describe which dataset and decision they were connected to. They also asked Meta to state under oath if it could identify no one. Hixon said the request was not limited to formal decisions.
Meta argued that its early AI dataset work followed an “informal approach,” without formal decisions by people with authority. It also contended that discussions were collaborative, meaning there was no single decision-maker. Hixon said the request covered all decision-makers, rather than assuming there was only one.
Staff accounts raise questions about approval
Meta has argued that staffers technically downloaded torrents, rather than anyone deciding to pirate books for training. But testimony and messages described in the court filings point to questions about oversight. One staffer said downloads “required some director or VP level approval to use for training.” Another testified that leadership halted and later resumed downloads sourced from Anna’s Archive.
Internal messages also showed workers pushing back on torrenting. One wrote that “torrenting from a corporate laptop doesn’t feel right.” The publisher argued that Meta’s position—that no one made a decision—does not account for repeated torrenting of shadow libraries from 2022-2024.
The plaintiffs have argued that torrenting involved both downloading and distributing files among users. They say Meta downloaded 80.6 terabytes from LibGen and 35.7 terabytes through Anna’s Archive, and contend that the scale and speed of the downloads indicate Meta was also seeding. Meta has said it used “mitigations” to prevent seeding.
Server records may show how transfers happened
Hixon’s order also requires Meta to produce “command history files for every server Meta used to torrent, including its virtual machines and AWS instances.” Such records could help the plaintiffs assess whether Meta seeded works and understand the company’s alleged mitigation methods.
That evidence matters to the plaintiffs’ contributory infringement claim, which the article describes as requiring them to prove that Meta facilitated transfers of their works. The order does not resolve the copyright claims. It requires additional information that could help establish what happened and who was involved.
Meta must answer more than a request for names
Meta had proposed providing names without details such as roles, approximate dates, or whether individuals contributed to, directed, permitted, or approved decisions. Hixon disagreed that names alone would be an adequate response. The plaintiffs had also asked about people who approved seeding, but Meta said it was not currently aware of anyone who made, authorized, or approved an upload or distribution decision.
The dispute includes early AI training for models like Llama 1 and 2. Meta said its initial informal approach later gave way to more structured review and approval processes. Hixon’s order requires answers that address who was involved even if decisions were informal or collaborative.
Meta must also explain what happened to pirated works that were not used for AI training. Taken together, the required disclosures could clarify the chain of decisions, the operation of the torrenting systems, and the handling of files beyond those used in training. The lawsuits will turn on evidence and legal arguments that remain to be resolved.