AirTag reveals how Amazon destroys rare books for AI training

Details

An investigation by 404 Media placed an AirTag inside a shipment of rare books and tracked it to an Amazon warehouse in Las Vegas, where a team called VGT3 (whose logo features a T-rex holding a book) cuts off book spines to accelerate scanning. Amazon confirmed it buys books through commercial channels to improve its products, specifically to train its Nova models. Booksellers believe AI companies are methodically working through entire catalogs by ISBN number. Anthropic ran a near-identical program called 'Project Panama,' exposed via a book-authors lawsuit. A judge ruled the practice constitutes fair use in part because originals were destroyed rather than copied and resold — legal reasoning that may inadvertently incentivize further destruction. The value of print books as training data comes from what they are not: they predate 2022, contain no AI-generated content, and much of the material has no digital equivalent. That scarcity makes them uniquely valuable as training data — and destruction of irreplaceable copies uniquely concerning. The practice concentrates knowledge that once sat on public shelves inside proprietary AI systems. The scale remains unknown, but the systematic warehouse-scale operation suggests it is far beyond incidental purchases.