Nvidia accused of using pirated books to train AI models

Nvidia is facing fresh scrutiny after court filings alleged that the company used pirated books and other copyrighted material from online shadow libraries to train its artificial intelligence models.

The claims were brought to light as part of the ongoing process of discovering facts for a pending class-action litigation case by authors in the United States.

“Nvidia, the AI chipmaker, appears to have greenlit a proposal made by their data strategy team to pay for high-speed access to Anna’s Archive, a well-known ‘shadow library’ search engine,” documents filed with the court appear to claim.

Court documents highlight that the company’s upper echelons were signing off on the scheme in a matter of days. It is alleged that the company committed copyright infringement with the help of datasets like ‘Books3’, which can be traced back to pirate sites like Bibliotik.”

According to the authors’ claim, Nvidia developed its AI models from various copyright-protected books without permission and allowed its corporate customers to access the data sets containing the infringing materials.

After the recent revelations, the plaintiffs filed an amended complaint in order to expand the case's scope. To defend itself against the recent allegations of copyright infringement involving its data sets, Nvidia has claimed fair use on its part.

What is Anna’s Archive?

Anna’s Archive is an open-source search engine that helps users locate books, academic papers, and other texts hosted on third-party sites. While it does not host pirated files directly, copyright holders frequently target its linked domains with takedown requests. The platform focuses largely on text-based material, which it describes as having high information density.