Authors Expand Nvidia Lawsuit Over Pirated AI Data

A group of prominent authors has expanded a class-action lawsuit against Nvidia, accusing the tech giant of deliberately seeking access to millions of pirated books and academic papers to train its artificial intelligence models. The amended complaint, filed on January 16 in the U.S. District Court for the Northern District of California, alleges that Nvidia executives authorized contacts with shadow libraries despite warnings about their illegal nature, amid intense competition in the AI sector.

The plaintiffs, including authors Abdi Nazemian, Brian Keene, Stewart O'Nan, Andre Dubus III, and Susan Orlean, claim Nvidia's data strategy team reached out to Annas Archive, a search engine aggregating content from various online libraries, including pirated materials. Internal emails cited in the filing show Nvidia inquiring about using the site's collection for pretraining large language models like NeMo, Retro48B, InstructRetro, and Megatron. Annas Archive reportedly warned Nvidia that its holdings were illegally obtained and sought confirmation to proceed. According to the complaint, Nvidia management approved the outreach within a week, leading to an offer of high-speed access...