AI/TLDR

Amazon · 2026-08-17 · major

Amazon destroys rare books to train AI — an AirTag traced the shipment

404 Media hid an Apple AirTag in a rare book and followed it to an Amazon warehouse in Las Vegas, where a team called VGT3 cuts the bindings off printed books so the pages scan faster. The printed copy is destroyed in the process.

Wall of bookshelves stacked with books
Studio 642 / Getty Images

A bookseller hid an AirTag in a rare book, and it ended up at an Amazon site that cuts books apart to scan them.

Quick facts

Reported by404 Media, August 17, 2026
FacilityAmazon warehouse, Las Vegas, Nevada (team VGT3)
Order traced1,000 books bought on Biblio
Tracking methodApple AirTag hidden between the pages
What happensBinding cut off, pages scanned, book destroyed
Amazon's statementBooks bought "through commercial channels"

What is it?

404 Media has documented where Amazon's bought books physically go. A bookseller who suspected that a 1,000-book order placed on Biblio — a marketplace where the buyer can stay anonymous — was really an AI company slipped an Apple AirTag between the pages of one volume. The tracker ended at an Amazon warehouse in Las Vegas run by a team called VGT3.

How does it work?

The pipeline is physical, not digital. Amazon buys printed books through ordinary commercial channels, ships them to the Las Vegas site, and staff there cut the bindings off so the loose pages run through a scanner much faster. The scan becomes text; the printed copy does not survive. Workers at the site describe the split simply — some people cut, some people receive and scan barcodes.

Why does it matter?

This is the first public look at how a major AI company sources training text off the internet, and it shows the cost is paid in physical books. Labs have largely used up the public web, so out-of-print writing is valuable precisely because it is not online and was written before models could write it. Booksellers and rights holders now have a concrete reason to ask who is behind an anonymous bulk order.

Who is it for?

booksellers, publishers and anyone tracking where training data comes from

Frequently asked questions

Which Amazon site receives the books?
The AirTag led 404 Media to an Amazon warehouse in Las Vegas, Nevada. The team working there is called VGT3 and uses a logo of a dinosaur holding a book in its claws. Staff describe the job in plain terms: they receive large shipments of printed books, some people cut the books, and others receive them and scan the barcodes.
What did Amazon say about buying the books?
Amazon told 404 Media that the company "purchases books through commercial channels to help develop and improve the products and services our customers use." Amazon did not name a specific model, product or training program in that statement, and did not say how many books it buys or what happens to the scans afterwards.
Why are old printed books valuable as AI training data?
Rare and out-of-print books are useful to model builders because the text is not already on the public internet, which labs have largely exhausted. Anything printed before 2022 also carries a near-certainty that a human wrote it, so it helps avoid model collapse — the decay that sets in when a model is trained on the output of other models.
Are the books destroyed for good?
Yes. 404 Media reports that workers at the Amazon site cut the bindings off so the loose pages feed through a scanner quickly, and the printed book does not survive that step. Because the shipments include rare and out-of-print titles, some of the physical copies going through the process are hard or impossible to replace.

Try it

https://www.404media.co/we-tracked-a-shipment-of-rare-books-it-ended-at-an-amazon-ai-training-facility/

Sources · 4 outlets

Tags

  • amazon
  • training-data
  • ai-training
  • data-provenance
  • books
  • copyright
  • publishing
  • model-collapse
  • investigation
  • 404-media

← All releases · Learn AI