• Home
  • Blog
  • Amazon Is Cutting Up Rare Books to Feed Its AI Machine

Amazon Is Cutting Up Rare Books to Feed Its AI Machine

Updated:August 17, 2026

Reading Time: 2 minutes
A stack of books
  • Home
  • Blog
  • Amazon Is Cutting Up Rare Books to Feed Its AI Machine

Amazon Is Cutting Up Rare Books to Feed Its AI Machine

A stack of books

Updated:August 17, 2026

Amazon built its empire on books. Now it may be destroying them. A new investigation reveals that the company is buying up rare books, slicing off their spines, and scanning every page.

Each page is fed to hungry AI models that need fresh material to learn from. 404 Media reporters planted a tracking device inside a rare book and watched where it went. 

The trail ended at an Amazon facility in Las Vegas.

Paper Trail

The Las Vegas facility has a name: VGT3, and a logo. A cartoon dinosaur clutches a book in its claws, almost like a mascot for this strange new chapter in publishing history.

Amazon confirmed their story. A spokesperson told 404 Media that Amazon “purchases books through commercial channels to improve the products and services customers use.”

That’s corporate speak for what’s really happening; Amazon needs massive data, and rare books are becoming one of the last untapped sources.

Rare Books

Books on shelves
Image Credits: Studio 642

AI models learn by reading massive amounts of text. Then they use patterns from that text to generate new writing.

The problem is, these models already devoured most of the internet. There’s not much fresh material left online. So tech companies are scrambling for new sources.

Books, especially old and out-of-print ones, contain knowledge that never made it onto the web. Nobody digitized them. Nobody typed them into a blog post. 

This isn’t just an Amazon problem, either. Anthropic, another major AI company, ran into legal trouble over this exact issue. 

The company settled a copyright lawsuit for 1.5 billion dollars after it was found to have used pirated books to train its models.

Model Collapse

AI models are starting to face a serious issue called model collapse. This happens when an AI trains on text that other AI wrote. 

Over time, the quality gets worse. Errors pile up, and the writing becomes repetitive and strange. Picture making a photocopy of a photocopy of a photocopy. 

Each copy gets a little blurrier. This is exactly why old books are so valuable now. Anything written before 2022 is guaranteed to be human-made. 

No chatbot could have written it. That makes these dusty old pages incredibly precious to companies racing to build smarter AI.      

Literary History

There’s something almost poetic, and a little unsettling, about this whole situation. Amazon started as a humble online bookstore back in the 1990s. 

Books built the company from the ground up. Now those same books are being physically destroyed. Their spines get cut, and their pages get scanned. 

Then, presumably, they get discarded once the data is extracted. Rare book collectors and historians might see this as a real loss. 

Once a book is destroyed, it’s gone. No amount of digital scanning can replace holding the original in your hands.