Amazon Trashes Rare Books for AI Training Data
A hidden tracking device confirms the tech giant is purchasing and destroying rare volumes to secure unique training data, sparking ethical debate and criticism from industry figures.
An investigation by 404 Media has revealed that Amazon is purchasing rare books and systematically destroying them at a Las Vegas facility to scan their content for artificial intelligence training. The discovery was made after a bookseller planted a hidden AirTag in a rare book included in a bulk order, which tracked to the VGT3 facility. Workers at the site are reported to tear books from their spines and scan pages, a practice confirmed by a logo of a Tyrannosaurus rex devouring a book displayed on the warehouse door.
Amazon declined to comment specifically on AI training but provided a statement that it purchases books through commercial channels to help develop and improve customer products. The move addresses a reported shortage of unique text data, with workers targeting books by ISBN to ensure comprehensive coverage. The practice has sparked ethical concerns among booksellers and the public regarding the loss of historical and intellectual value.
The VGT3 facility in Las Vegas houses a team specifically focused on destructive book scanning for AI training. Amazon reportedly faced a shortage of unique text data earlier this year, with workers expressing concern that the warehouse might shut down if more books were not found. Workers at the facility were trained to scan barcodes or ISBNs before scanning books, supporting the theory that AI firms are methodically scanning every printed book by working through ISBN lists.
The books targeted are often older, lower monetary value items, such as untranslated foreign language books or poorly distributed titles, rather than prized first editions. Michael Burry has reportedly described the practice as “evil incarnate,” while Scottish bookseller Derek Walker has called for AI firms to distinguish between texts that won’t be missed and unique surviving editions.
AI firms are in a race to advance frontier models, requiring massive amounts of unique training data to stay competitive with rivals like Google, OpenAI, and Anthropic. Rivals such as Anthropic and xAI have publicly stated they are not training on rare or antique books. Booksellers have long suspected that AI firms were buying up rare books for destruction, but proof was difficult to obtain until the AirTag investigation.
The practice of destructive book scanning has sparked ethical debates, with critics arguing it destroys historical and intellectual value, while some online commentators suggest the books were gathering dust and that AI could make knowledge more accessible. Amazon is developing frontier AI models and guards its training data to avoid losing a competitive edge.
