US Judge Allows New Contributory Infringement Claim in Authors vs. Meta — BitTorrent Seeding at Issue
Judge Vince Chhabria of the Northern District of California 'reluctantly' granted a motion by lead plaintiff Richard Kadrey to file a fourth amended complaint against Meta, adding a contributory infringement claim related to Meta's use of BitTorrent to acquire approximately 81.7 terabytes of pirated books — over 7 million titles — from shadow libraries including LibGen and Z-Library. Because BitTorrent requires users to 'seed' (upload) data to others while downloading, the authors argue Meta contributed to third-party copyright infringement. The judge granted the motion to protect absent class members but sharply rebuked the plaintiffs' lawyers at Boies Schiller Flexner for a 'pattern' of blaming Meta for their own litigation failures.
Analysis
The Kadrey v. Meta case has now produced its most legally interesting development yet — and the most revealing one about the state of AI copyright litigation strategy.
Judge Chhabria's June 2025 ruling that Meta's use of copyrighted books for AI training was protected by fair use appeared, at the time, to be a significant victory for the tech industry. But the case did not end there, because the fair use ruling addressed what Meta did with the books, not how it obtained them. The contributory infringement theory now being added to the complaint addresses the second question: by using BitTorrent to download pirated books, and by necessarily uploading ("seeding") fragments of those files to other users in the process, did Meta contribute to the copyright infringement of third parties?
This is a genuinely novel legal theory in the AI training context, and it is structurally distinct from the direct infringement claims that have dominated AI copyright litigation. Contributory infringement does not require proving that Meta's AI training was infringing — it requires proving that Meta's data acquisition method facilitated infringement by others. The BitTorrent protocol's seeding requirement makes this argument unusually concrete: every byte that Meta downloaded from a shadow library was also, by technical necessity, uploaded to other users of that torrent swarm.
The judge's rebuke of Boies Schiller Flexner is worth noting not just for its colourful language — "lame excuse," "false statements," "pattern of blaming Meta" — but for what it reveals about the litigation dynamics. The plaintiffs' legal team has been aggressive and, according to the judge, sometimes reckless in its approach. The fact that the judge granted the motion despite his frustration with counsel, specifically to protect absent class members, suggests that he views the underlying legal theory as having sufficient merit to warrant adjudication even if the lawyers presenting it have not covered themselves in glory.
The parallel with the Bartz v. Anthropic settlement is direct. Both cases converge on the same legal principle: the source of AI training data, not the training itself, is where the most durable copyright liability lies. The Anthropic settlement established the price of that liability. The Meta case will determine whether it can be extended through the contributory infringement theory to cover the act of acquiring pirated data, not just the act of training on it.