So to get to the bottom of the mystery, they agreed to place an Apple AirTag between the pages of one of the volumes. The tracking device traveled across multiple US states, eventually terminating at a high-security Amazon facility in Las Vegas, Nevada.
Inside this warehouse, a specialized unit code-named VGT3 systematically processes incoming books by slicing off their bindings, scanning the pages for digital intake, and destroying the physical originals.
This aggressive scanning pipeline highlights the extreme lengths to which tech companies will go to secure high-quality training datasets, particularly pre-2022 literature that is completely free of AI-generated content.
While independent booksellers watch their inventories get shredded, legal and ethical questions are mounting over the practice of turning physical, copy-protected books into digital training data before disposing of the evidence.
Amazon defended the operation in a brief statement, claiming they purchase books through commercial channels to improve their products and services, but the revelation has ignited intense debate over the systematic destruction of physical human knowledge to fuel artificial systems.
Futurism's Frank Landymore reported on August 17, 2026 that the public only became aware of this AI industry practice in a lawsuit against Anthropic, which revealed that the Claude chatbot maker was using industrial equipment to cut the pages out of millions of books and scan them, before they were disposed of.
More alarming was the upshot of that lawsuit, Bartz vs. Anthropic when Judge William Alsup in the US District Court for the Northern District of California ruled on June 23, 2025 that this specific one-for-one digital conversion of legally purchased print books for internal AI training was "spectacularly transformative" and constituted fair use, as the digital copies were not shared or sold outside the company.
Since then, book sellers have often suspected that their wares are being used, and then destroyed, to train AI models. In previous 404 reporting, one said that some of the giveaways were the size of the orders, the seemingly random selection of books, and that the books all had ISBNs. But they couldn’t prove this hunch because the orders are anonymous.
As 404 notes, that the workers at the Amazon site are apparently scanning the ISBNs of all the books they process gives credence to a spooky theory by booksellers: that AI companies are trying to scan every printed book in the world by going through a list of all their serial numbers.
The bookseller 404 worked with sold books that are rare, meaning that there are few copies left in circulation. That can be for a myriad of reasons – they’re not necessarily first editions of old classics – but they’re still valuable.
“There are different types of value”, the bookseller told 404. “There’s monetary value, obviously, but there are a lot of other types of value. There’s historical value, intellectual value, sentimental value. All sorts of things, and all of those the AI companies don’t care about. They just want the content as a bunch of words strung together”.





































