Amazon’s Las Vegas Warehouse Reportedly Cuts Up Books for AI Training Data
A tracked shipment and a warehouse employee’s account connect Amazon’s VGT3 facility to a process that captures book pages digitally while permanently destroying the bound copies.
Listen to this story
The audio brief
Story brief
3 key pointsA 404 Media investigation links Amazon’s VGT3 facility in Las Vegas to a reported book-digitization operation that destroys bound volumes after scanning them. A former worker described at least 20 high-speed scanners processing new, used, library-liquidated and foreign-language books, while another account involved rare books tracked to the warehouse. The reporting does not establish whether Amazon or an outside...
- 01
A VGT3 worker estimated 20–25 scanners were operating on the warehouse floor.
- 02
The reported workflow removes bindings, scans loose pages, then discards the physical books in mixed piles.
- 03
Materials reportedly included German, Russian and Japanese books, plus volumes associated with London and the University of London.
Amazon’s VGT3 warehouse in Las Vegas is reportedly destructively scanning books for AI training data, using a workflow that cuts off bindings, scans loose pages and discards them in mixed piles. The reported result is a digital copy of the text and a physical book that cannot be reassembled.
The account began with a shipment of rare books. 404 Media said it placed a tracking device in books a seller suspected were being bought by an anonymous AI company, then followed the shipment across the country to VGT3. The publication also found online accounts from people who said they worked there and described receiving books, cutting their spines and scanning the separate pages.
From bound book to loose sheets
An anonymous VGT3 employee interviewed by 404 Media described books arriving in boxes or bulk containers before workers moved them toward cutting stations. At each station, an operator slides a book into a guarded machine, removes their hands and activates a blade that removes the spine. The separated pages are then fed through high-speed scanners and later thrown into large cardboard containers, the employee said.
The employee said the incoming material covered far more than recent commercial titles: new and used books, items apparently liquidated from libraries, books from London and the University of London, and volumes in German, Russian and Japanese. They also recalled documents presented to Parliament. Those descriptions point to a broad intake stream, though they do not establish the rarity, ownership history or rights status of individual books.
The employee said workers were initially told the project was for Kindles, but later concluded it was for AI training. 404 Media identifies VGT3 as an AI-training-data operation. It does not identify a downstream model or establish whether Amazon was collecting the scans for itself or another company—an important unanswered question about who ultimately uses the material.
Sources
- 404media.coInside the Warehouse Where Amazon Scans and Destroys Books for AI Training
