Amazon Buys Old Books, Scans Them for AI and Destroys the Originals

Amazon is quietly purchasing vast quantities of used books, stripping their bindings, and digitizing the pages to feed its artificial intelligence models before destroying the physical copies.

The operation, uncovered by a 404 Media investigation, was exposed after a bookseller placed an AirTag inside a volume sold through the online marketplace Biblio.

The tracking device traveled across the U.S. before arriving at Amazon’s LAS8 facility in North Las Vegas.

Inside the warehouse, an internal unit dubbed "VGT3" orchestrates the rapid digitization process. Employees at the facility systematically sever book spines to accelerate mass scanning.

The Data Squeeze

“Amazon purchases books through commercial channels to help develop and improve the products and services our customers use,” an Amazon spokesperson stated, declining to specify the scale of the acquisitions.

The aggressive procurement strategy highlights a looming crisis for the tech sector: the depletion of human-authored text available for machine learning.

Analysts at Epoch AI project that developers could exhaust the existing supply of high-quality public data as early as 2026.

To bridge this gap, tech giants are targeting physical publications that predate the generative AI boom or lack a digital footprint.

Independent booksellers have reported a surge in massive, indiscriminate orders for obscure titles with ISBNs dating back decades.

Industrial Digitization

These bulk buyers operate under opaque monikers like the Blue Finch Project and Red Sparrow Project.

They buy hundreds of unrelated titles at a time, route them through logistics hubs, and never negotiate price.

Destructive scanning is not unique to Amazon. Anthropic previously admitted to buying millions of physical books, scanning their contents, and discarding the remains to build its own digital research library.

This method of "format shifting" has faced legal scrutiny, though a U.S. federal judge recently ruled that Anthropic’s one-to-one conversion constituted fair use because it did not introduce new copies into the commercial market.

Preservation Concerns

The industrial-scale destruction of inventory has alarmed archivists.

While organizations like the Internet Archive photograph texts without damaging bindings, corporate tech firms opt to slice spines strictly to reduce operational costs and maximize speed.

Some antiquarian dealers worry the indiscriminate buying could wipe out the last remaining copies of obscure works.

“If we’re going to save the signed copies, we have to sacrifice the unsigned,” noted one Portland bookseller who recently intercepted an order destined for a scanning facility to rescue a rare inscribed edition.