404 Media traces rare books to an Amazon AI training facility
An AirTag hidden in a purchased book led to a Las Vegas facility, internally called VGT3, where workers reportedly cut spines and discard the books after scanning them.
- Culture & impact
- Courts & copyright
- Notable
404 Media reported that it traced a shipment of rare books to an Amazon warehouse in Las Vegas where staff scan printed books and discard them, after a bookseller on the marketplace Biblio noticed a run of unusually large, thematically unrelated orders and suspected the buyer was assembling an AI training set. Reporter Emanuel Maiberg hid an Apple AirTag inside one volume of a roughly 1,000-book order and tracked its route to the site.
The tracker led to a unit that identifies itself internally as VGT3, marked with a logo of a dinosaur holding a book in its jaws. According to accounts from workers there relayed to 404 Media, incoming shipments have their spines cut off so the loose pages can be scanned more quickly, after which the printed originals are discarded; the scanned text is used to train Amazon’s AI models. Amazon’s on-the-record response to 404 Media was that it “purchases books through commercial channels to help develop and improve the products and services our customers use” — a statement that did not address what happens to the physical books, and which other outlets reported ran counter to an earlier company denial that it destroyed books during scanning.
The practice echoes a detail that surfaced the previous year in Anthropic’s book-copyright litigation, where court filings described the company buying print books, scanning them and discarding the copies — a process a federal judge held was fair use when applied to lawfully purchased books. Both Anthropic and xAI have since said publicly that their own book-acquisition efforts avoid purchasing and destroying rare or antiquarian volumes; Amazon offered no equivalent assurance.
The investigation added a company known first as a bookseller to the list of AI developers whose data-sourcing practices have drawn scrutiny, and demonstrated how little outsiders can otherwise verify about what happens to purchased data once a company’s public statement is limited to naming a “commercial channel.”