AirTag planted in a rare book led investigators straight to Amazon’s hidden AI scanner
Amazon is accused of buying up thousands of potentially 'rare' books, scanning them to train artificial intelligence, and then pulping them. While Amazon has since massively expanded into selling everything from robot dogs to jet skis, remember its humble roots as an online bookstore running out of a rented garage.
There are continued reports that companies like Amazon, Anthropic, and xAI are buying tons of old books and scanning them to train AI models, with previously revealed court documents seeming to shine a light on this practice. The problem is that copyright law calls the whole situation into question, potentially requiring the original text to be destroyed.
Now, an investigation by 404 Media involved co-founder Emanuel Maiberg having an Apple AirTag in a supposedly rare book and seeing where it went.
Amazon accused of buying rare books to train AI
Maiberg maintains that an anonymous bookseller informed him about a shipment for around 1,000 books that was made through the Biblio marketplace. They agreed to slip a 404 Media Apple AirTag into a book that they "suspected would be acquired by an AI company for training data" and tracked it from California to Milwaukee, before it ended up in an Amazon warehouse in Las Vegas known as LAS8.
Advert
Although LAS8 is known as a print-on-demand book company, the AirTag was tacked to the north end of the warehouse, which Amazon employees on a workers' forum claim is a different operation known as VGT3.
Complete with the logo of an angry Tyrannosaurus rex clasping a book, VGT3 is said to be split into those who cut books and those who scan barcodes. With the spines sliced off, it's alleged that the pages of books are scanned for AI training, with critics complaining it's not even to digitize these materials for preservation.
The BBC recently reported on this exact issue, explaining how there's been a recent boom in second-hand book sales amid suspicions that they're being fed to AI and then destroyed. In September 2025, Anthropic agreed to settle for $1.5 billion after authors accused it of 'stealing' their work to train its models.
The settlement came just two months after Judge William Alsup ruled that Anthropic's use of books was "exceedingly transformative" and allowed under US law. Anthropic ended up falling down over claims that it violated the rights of authors by saving over seven million pirated copies of their works in a "central library of all the books in the world."
The AI giant settled as it faced up to $150,000 in damages for every copyrighted work.
Why could AI companies be destroying books?
Putting older and potentially rare books in the crosshairs, those printed before 2022 are said not to have AI-generated text, meaning models won't be subjected to ingesting AI materials in a process known as 'model collapse'.
Documents in the Anthropic case refer to internal communications about 'Project Panama', which is part of the company's aim to destructively scan all the books in the world."
As for why books are being destroyed, University of California copyright law professor Pamela Samuelson referred to it as 'format shifting' as she explained why Anthropic would destroy a physical copy to make a digital replica: "It [Anthropic] didn’t multiply the number of copies of the work in existence so not an infringement."
Amazon has previously denied that it takes part in 'destructive scanning', although the 404 Media bookseller warned that the kind of books heading the company's way could be rare for a number of reasons: "There are different types of value. 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.”
In response to 404's latest claims, Amazon stated: "Amazon purchases books through commercial channels to help develop and improve the products and services our customers use."
The jury is out on the practice in general, with Derek Walker, owner of Edinburgh bookshop McNaughtan's, telling the BBC that while the loss of one copy of a recent academic text that already has numerous copies in a library won't be missed. Still, Walker admitted: "But we have, and have sold, books which are for example the only known surviving example of an edition from the 18th century.
"It would be a much more significant problem if one like that were to be bought for destruction, having survived this long."