We Tracked a Shipment of Rare Books. It Ended at an Amazon AI Training Facility
Amazon is systematically acquiring and destroying rare books for AI training — and nobody knew.
Why it matters
This is a work/labor/society story about how AI training data sourcing is reshaping what happens to cultural artifacts and intellectual property. It raises questions about copyright, fair use, and the hidden costs of frontier model training that practitioners and companies need to understand.
The key facts
10 to knowTracking investigation placed device in rare book shipment
Shipment traced to Amazon facility dedicated to AI training data
Books being scanned and destroyed at scale
Reveals hidden sourcing practices in generative AI model training
Raises copyright and intellectual property concerns
Amazon operating a facility that scans and destroys books for AI training
Rare books being acquired and used without apparent author/publisher consent
Tracking device investigation by 404 Media reveals the supply chain
Raises fair-use and copyright implications for AI training data sourcing
Signals potential regulatory/legal pressure on tech companies' data practices
Go to the source
404 Media404media.co
Publisher excerpt: We placed a tracking device in a shipment of rare books to see which AI company was buying it, and found an Amazon facility where Amazon scans and destroys books.