Amazon is Destroying Rare Books to Secretly Train its AI Models

An investigation by 404 Media has exposed a hidden operation inside Amazon. The tech giant is found to be buying up massive quantities of rare physical books, slicing off their spines to speed-scan the pages and feed its Nova AI models before destroying the originals.

The tracking operation started when reporters partnered with a bookseller who noticed unusually bulk orders for rare titles. They slipped an Apple AirTag inside a shipment of 1,000 texts and watched its journey across the United States.

The package made its way to a massive Amazon facility in Las Vegas, Nevada, known as LAS8. Inside, a specific division operating under the code name VGT3 handles the processing. Fittingly, the logo on the team’s entrance features a Tyrannosaurus Rex brandishing its teeth while holding a book.

Inside the facility, the physical media meets a grim end. Workers at the warehouse report that their day-to-day duties consist of taking bulk shipments of books, slicing their bindings off to run pages through high-speed scanners, and discarding the destroyed remnants. The digitized text is used to feed data into Amazon’s proprietary AI models, such as Nova.

When questioned about the operation, an Amazon spokesperson offered a brief statement: “Amazon purchases books through commercial channels to help develop and improve the products and services our customers use.”

The practice highlights a growing desperation among tech giants to find high-quality text for machine learning datasets. As the open internet becomes saturated with low-quality, AI-generated content, pre-2022 physical print media has become a goldmine for tech companies. Older books offer human-written grammar, niche information, and clean prose untouched by digital slop.

Amazon is not the only company utilizing this destructive approach. A previous lawsuit involving AI startup Anthropic revealed a similar initiative dubbed “Project Panama,” where the company bought printed books on secondary markets, sliced off their spines, and scanned them.

In that legal case, a judge ruled that destroying the physical original after scanning actually helped shield the company under fair use laws, as the physical copies were not being resold or distributed directly.