Amazon Is Buying and Destroying Rare Books to Train AI

A rare bookseller suspected AI companies were quietly buying up rare books to train their models, then destroying them. To find out for sure, they planted an AirTag inside a shipment and tracked where it went.

It led to a warehouse in Las Vegas called VGT3, part of Amazon, according to a 404 Media investigation. Workers there cut the spines off incoming books to scan pages faster, destroying the original in the process. The team’s logo shows a Tyrannosaurus rex devouring a book.

Amazon wouldn’t confirm the books are being used for AI training. Its statement only said the company “purchases books through commercial channels to help develop and improve the products and services our customers use.” But Amazon is building competitive frontier AI models that need massive, unique training data, and workers reportedly said the facility nearly shut down earlier this year after running out of books to scan.

404 Media also found evidence supporting a theory that AI firms are systematically working through lists of ISBNs to make sure every unique book gets scanned, Ars Technica reported. Workers said they’re trained to check barcodes before scanning.

Not every rare book is worth a fortune, but booksellers say many still carry real historical or sentimental value, the kind of value AI companies “don’t care about,” one told 404 Media. “They just want the content as a bunch of words strung together.”