NVIDIA accused of using pirated books to train AI

A lawsuit claims NVIDIA knowingly used millions of illegally copied books to develop its artificial intelligence models.

According to a report by TorrentFreak, a lawsuit alleges that NVIDIA leaders approved communication with Anna’s Archive, a website providing free access to a vast collection of copyrighted books and research papers.

Leaked NVIDIA emails suggest someone from their data strategy team contacted Anna’s Archive to see if the site’s data could be used to train large language models.

NVIDIA accused of using pirated book site

According to TorrentFreak, the updated lawsuit argues that NVIDIA turned to piracy because of competition. It claims the company deliberately sought out illegally obtained book data as the need for information to train its AI systems grew.

The updated lawsuit also claims NVIDIA used illegal sources to obtain copyrighted books, specifically mentioning downloads from LibGen, Sci-Hub, and Z-Library.

The lawsuit states that Anna’s Archive informed NVIDIA the data it possessed was obtained illegally and asked if NVIDIA still wanted to move forward. According to the complaint, NVIDIA approved the use of the data within a week, and Anna’s Archive then reportedly offered NVIDIA access to around 500 terabytes of it.

The documents also claim that some of the content shared was typically accessible only through the Internet Archive’s lending program, which is currently involved in several copyright lawsuits.

In addition to directly providing the service, NVIDIA is accused of sharing programs and resources that allowed businesses to automatically obtain datasets filled with illegally copied books.

NVIDIA has maintained that its use of data to train AI models, including NeMo, is legally permissible under fair use principles and doesn’t violate copyright laws.

Anna’s Archive has faced scrutiny from major tech companies before. Last December, Spotify investigated reports that the site had collected and uploaded 300 terabytes of its data.

Read More

2026-01-20 16:49