Amazon destroys rare books for AI training data

Aug 17, 2026

An investigation by 404 Media revealed that Amazon is purchasing rare and out-of-print books, cutting off their bindings, and scanning them to train its Nova AI models. The books are destroyed in the process. AI companies target pre-2022 printed texts to avoid model collapse from AI-generated online content. While a prior ruling on Anthropic's similar 'Project Panama' deemed the practice fair use because originals were destroyed, critics warn it permanently locks irreplaceable public knowledge inside private corporate models.

Amazon book scanning operation

▪Amazon employees at the Las Vegas facility, known as VGT3, cut the bindings and spines off printed books to scan them more quickly, destroying the books in the process.
▪Amazon uses the scanned book data from its Las Vegas facility to train its Nova artificial intelligence models.
▪An investigation by 404 Media tracked a shipment of rare books containing a tracking device to an Amazon facility in Las Vegas, Nevada.
▪Amazon stated that the company purchases books through commercial channels to improve the products and services that customers use.

AI training data sourcing

▪Booksellers suspect that artificial intelligence companies are attempting to systematically scan every printed book by its International Standard Book Number.
▪Rare and out-of-print books offer a valuable source of training data because they often do not exist on the internet.
▪Artificial intelligence companies seek printed books published before 2022 because they are free of AI-generated text, which can cause model collapse.

Physical book destruction controversy

▪The practice of destroying original printed books for scanning is controversial because it removes potentially irreplaceable knowledge from public shelves.
▪Critics argue that digitizing and destroying rare books locks public knowledge inside the closed artificial intelligence models of a single corporation.

Anthropic Project Panama precedent

▪A lawsuit by book authors revealed that Anthropic operated "Project Panama," which involved purchasing books on marketplaces, cutting off their spines, and digitizing them.
▪Anthropic has previously ingested illegally pirated books from the internet to train its large language models.

Fair use legal ruling

▪A judge ruled that Anthropic's digitization of books qualified as fair use and did not violate copyright laws.
▪The legal ruling in the Anthropic lawsuit was partly based on the fact that the printed originals were destroyed rather than copied and resold.