GoKawiilFormer Meta Researcher Testifies He Torrented Pirated Books3 Dataset
Stephen Roller, who worked at Meta from 2017 to 2023, testified in a deposition that he downloaded the pirated Books3 collection via BitTorrent and knowingly distributed it while employed there, according to rightsholders suing Meta. He said he later deleted his copy because of ongoing litigation, though he could not recall the exact download date. Meta has disputed his account, maintaining it does not believe it obtained Books3 through torrenting.
GoKawiil's interpretation of the reporting above, not reported fact.
A federal judge already ruled that training Meta's Llama models on books counts as fair use, so the torrenting method of acquisition remains the main unresolved legal question in the authors' lawsuits. If Roller's testimony is credited by the court, it could undercut Meta's public denials and strengthen plaintiffs' arguments that Meta knowingly used pirated material obtained through illicit distribution networks. The case could also pressure Meta to disclose which employees approved sourcing from shadow libraries, a question a judge has already ordered the company to answer.
- Ex-Meta researcher Stephen Roller testified he torrented and distributed the pirated Books3 dataset while at Meta.
- Meta disputes this, saying it doesn't believe it obtained Books3 via BitTorrent, despite admitting it trained models on the dataset.
- A judge has ruled training on books is fair use, leaving the torrenting method as the key remaining legal issue in the authors' lawsuits against Meta.
Source: torrentfreak.com, 2026-10-31
Published there as: “Ex-Meta Researcher Torrented Pirated Books, Deleted It 'Due to Litigation'”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.