Amazon built its name selling books. Now, a trail of rare volumes reportedly ending up at one of its warehouses has raised the darker possibility that irreplaceable print is being consumed to feed artificial intelligence.

The alarm sharpened after 404 Media placed a tracker inside a shipment of rare books. The device ultimately led to Amazon’s VGT3 facility in Las Vegas, a warehouse identified by a dinosaur-with-a-book symbol.¹

Workers at the location told the outlet that the operation receives huge shipments of printed books, removes their bindings and scans them at speed. “The printed book is destroyed in the process.”² The account follows earlier reporting that AI companies have been buying used books in bulk for training data.²

The apparent logic is brutally simple: large language models require vast stores of text, while material freely available online has become increasingly exhausted, contested or polluted by AI-generated writing. Rare and out-of-print books can offer text that is difficult to find digitally—and, crucially, predates the current wave of generative AI.¹

Amazon did not address the reported destruction of books in the material provided. But the company told 404 Media that it “purchases books through commercial channels to improve the products and services customers use.”¹

That answer captures the divide. For Amazon, the books are commercially obtained inputs for better products. For critics of the practice, the tracker experiment suggests a one-way conversion of scarce physical culture into machine-readable data: bindings cut, pages scanned, and editions that may be impossible to replace reduced to scrap.¹