Background
Entrepreneur Media, publisher of Entrepreneur magazine and related digital content, sued Meta Platforms alleging that Meta obtained copyrighted works through BitTorrent file-sharing networks to use as training data for its large language models (LLMs). Unlike many AI copyright cases that focus on whether the training itself constitutes infringement, this case zeros in on how Meta acquired the training data.
The complaint alleges that Meta systematically downloaded copyrighted works from torrenting networks—a process that, by the nature of the BitTorrent protocol, requires the downloader to simultaneously upload (“seed”) portions of the files to other users on the network. Entrepreneur argues that this uploading constitutes both direct copyright infringement (unauthorized reproduction and distribution) and contributory infringement (enabling further infringement by other torrent users).
Meta moved to dismiss, arguing that the claims were insufficiently pleaded and that the alleged conduct fell within permissible uses.
The Court’s Holding
Judge Chhabria denied Meta’s motion to dismiss, finding that Entrepreneur adequately alleged that Meta’s use of torrenting networks to obtain copyrighted works “necessarily required uploading those works onto the torrenting network, thus enabling copyright infringement by third parties.” The court held that this theory of liability—focused on the acquisition method rather than the training use—was sufficiently pleaded to survive dismissal.
The ruling distinguishes this case from other AI training copyright disputes, such as Kadrey v. Meta, where Meta prevailed on fair use grounds. Here, the legal theory does not depend on whether training an LLM is fair use—it targets the separate act of distributing copyrighted works through a peer-to-peer network to acquire them in the first place.
Key Takeaways
- Copyright plaintiffs may have a stronger litigation path by targeting how AI companies acquire training data rather than how they use it—sidestepping the complex fair use analysis that has favored defendants in other AI training cases.
- The BitTorrent protocol’s inherent upload-while-downloading architecture creates a distinct copyright liability exposure: companies that torrent copyrighted content are simultaneously distributing it, even if their goal is only to download.
- This ruling does not address whether AI training itself infringes copyright—it leaves that question for another day. But it opens a separate front that could prove harder for AI companies to defend.
- The case may incentivize AI companies to document and clean up their training data acquisition pipelines, as the method of collection itself can create liability independent of the end use.
Why It Matters
This case opens a significant new front in the AI copyright wars. While most AI training lawsuits have focused on whether using copyrighted works to train models constitutes fair use, Entrepreneur Media’s theory attacks the supply chain—how the training data was obtained in the first place. If Meta indeed used BitTorrent to build training datasets, the inherent architecture of peer-to-peer file sharing means Meta was also distributing those works to others, a much harder act to defend under copyright law. For AI companies broadly, the ruling signals that the legality of training methods cannot be evaluated in isolation from the legality of data acquisition. And for content creators, it provides a potentially more powerful legal theory that does not require overcoming the fair use defense that has shielded AI companies in other cases.