AllegedThe record is a lawsuit, complaint or accusation that has not been decided.
Nvidia Accused of Training AI Models on Pirated Books
A class action lawsuit alleges that Nvidia trained its AI models on pirated books obtained from sources including Anna's Archive without authorization or compensation, and that internal emails show the company considered purchasing access to millions of pirated titles despite warnings about illegality.
19 January 2026 · Nvidia · Intellectual property and appropriation
A class action lawsuit filed in the United States accuses Nvidia of using pirated books without authorization or compensation to train AI models including NeMo, Megatron, and Retro.
The amended complaint alleges that Nvidia contacted Anna's Archive, a major piracy library, and considered purchasing high-bandwidth access to millions of pirated books despite receiving explicit warnings about the illegality of the contents. The complaint cites internal emails as evidence and references use of other pirate sources including Books3, LibGen, Sci-Hub, and Z-Library.
Reported impact
The case raises questions about whether fair use exceptions protect AI training on copyrighted works, particularly when pirated materials are involved.
Index analysis
No court has yet validated the internal emails presented by plaintiffs as evidence. Nvidia has not responded substantively to the new accusations and has previously defended its practices by invoking fair use protections, arguing that books constitute only statistical datasets for AI models.