AI companies are scanning truckloads of books—sometimes rare ones—and often destroying them in the process to train their models. A new investigation offers the first public picture of which companies are involved and how the operation works.

Independent tech media publication 404 Media placed a tracking device inside a shipment of rare books and watched it travel to an Amazon facility in Las Vegas, where the company scans and destroys printed books to train AI. The investigation, published Monday, is the first to publicly pinpoint where bulk book purchases tied to AI training end up.

The shipment's final stop was Amazon's VGT3 team. Employees told the outlet that their work consists of receiving massive shipments of printed books, then cutting off the bindings so the pages feed a scanner more quickly. The book itself is destroyed in the process. The team's logo—a dinosaur clutching a book in its claws—strikes an unusually fitting note.

Why printed books are suddenly in demand

Books printed before 2022 carry text that is mostly not readily available online and, crucially, is not machine-written. Training a model on its own kind of output risks "model collapse," a quality-degrading spiral researchers demonstrated in a 2024 paper published in Nature. Pre-2022 paper has therefore become clean fuel for AI training.

Amazon isn't the only one. A cottage industry has formed to supply AI labs with physical books that get stripped, scanned, and discarded—a real-life "Fahrenheit 451" scene of texts destroyed to feed machines. Anthropic's internal "Project Panama" already ran such an operation at scale, digitizing millions of books through destructive scanning. The method marks a break from the last era of mass digitization: Google Books scanned millions of library volumes in the 2000s and 2010s without consuming them—borrowed copies went back to the shelves—and a US appeals court ruled in 2015 that the project's scans were fair use.

The scramble for clean text has already landed in court. US District Judge William Alsup ruled that training AI on legally purchased books can qualify as fair use, a partial win for Anthropic that Meta and OpenAI also claimed. Pirated copies are a different matter: Anthropic agreed to a $1.5 billion settlement over scanned pirated titles, and Salesforce now faces a class action over alleged book piracy. Rights holders are pushing back in other ways, too: Penguin Random House, the world's largest trade publisher, added language to the copyright pages of its books in late 2024 prohibiting their use for training AI systems.

The Amazon operation stayed quiet for a structural reason. Booksellers noticed a spike in bulk orders over the past year and suspected AI companies were behind them—the buyers weren't price-sensitive and picked titles seemingly at random. Their concerns were finally confirmed by an Apple AirTag hidden in a major book order, which revealed the destination: Amazon's VGT3 warehouse in Las Vegas.

Amazon confirmed the operation hadn't been reported before, and said it "purchases books through commercial channels to help develop and improve the products and services our customers use." The company is also building AI products of its own: Amazon launched its Nova family of foundation models in December 2024 and sells AI services through AWS, businesses that depend on training data at scale.