🕒 Created · Updated

Amazon is Destructively Scanning Rare Books to Train AI Models

Amazon is systematically purchasing and destroying physical books to train its frontier AI models. An investigation by 404 Media revealed that the company uses a dedicated warehouse in North Las Vegas, labeled VGT3, to scan books at an industrial scale. Workers at the facility tear books from their spines and scan the pages, eventually recycling the remains. This practice, known as destructive scanning, allows Amazon to access unique training data from rare, antique, and out-of-print books that exist only in physical form. By purchasing the books legally, Amazon avoids copyright issues and can use the text as a transformative fair use for internal model training. While some critics, such as Michael Burry, have labeled the practice "evil incarnate," others argue that it is a necessary step for AI development. Booksellers worldwide have reported a surge in bulk orders for seemingly random titles, ranging from obscure Latin texts to cowboy novels, as tech giants race to feed their AI machines with human-authored content.

Sources