Rare Books Traced to Amazon AI Training Facility to Be Scanned and Destroyed
AI companies are buying physical books, sometimes rare ones, and destroying them during destructive scanning to obtain “clean” training data. A 404 Media investigation tracked a rare-book shipment to an Amazon facility in Las Vegas, where the VGT3 team reportedly removes bindings and scans pages faster, rendering books unusable. The process matters because pre-2022 books contain text less likely to be online or AI-generated, reducing risks of model collapse from training on synthetic content. Amazon is part of a wider supply chain serving AI labs, including Anthropic’s large-scale “Project Panama” book digitization efforts. The practice has also triggered legal fights: a judge said training on legally purchased books can qualify as fair use, while piracy-related scanning has led to a $1.5 billion Anthropic settlement and a class action against Salesforce. Amazon says it buys books through commercial channels to improve products and services.
