Nvidia may have “borrowed” millions of books from this shadow library for its AI

Photo: The interior of the Stockholm Public Library in Sweden. Photo credit: Tamás Mészáros via Pexels. Article by Konstancija Gasaitytė. Cybernews – January 20, 2026.

Authors who filed a lawsuit against the company have dug deeper to uncover Nvidia’s methods for obtaining new content to train its AI models.

Nvidia allegedly used millions of pirated books from Anna’s Archive to train its AI models. According to internal Nvidia documents cited in a class-action lawsuit, the company reached out to Anna’s Archive to access its data. […]

Click here to view original web page at www.cybernews.com

Related articles:

Judge orders Anna’s Archive to delete scraped data; no one thinks it will comply – Ars Technica

Anna’s Archive has ‘backed up Spotify’ and created a series of bulk torrents – BetaNews

Nvidia allegedly greenlit the use of pirated books from illegal sources to train its AI models, according to an expanded class-action lawsuit – PC Gamer