Suspicious bulk orders of used books have raised fresh concerns that AI companies are buying physical books, scanning their contents, and using the text for AI training without paying authors or asking for permission. A new investigation appears to give booksellers their strongest evidence yet after an AirTag hidden inside a rare book ended up at an Amazon AI training facility.
Used-book sellers have reported unusually large orders containing titles with little or no connection to each other. That pattern led some sellers to suspect that buyers were collecting books for large-scale scanning rather than for resale or personal use.
404 Media placed an AirTag inside a shipment of rare books to track where the order went, and the device eventually reached an Amazon facility where workers reportedly scan and destroy books.
Don’t miss the best of The Mac Observer
Set us as a preferred source and our Apple reporting ranks higher in your Google Search results and Discover feed — one tap, no account changes.
Employees at the facility said large shipments arrive before staff remove the bindings and scan the pages at high speed. The process destroys the original books, which allows workers to handle large volumes of printed material more efficiently.
Why used books matter for AI training
Older books have become especially valuable for AI training because material published before the recent AI boom is less likely to contain text generated or heavily edited by AI systems. That reduces the risk of training newer AI models on content already produced by other AI tools.
The practice has also raised copyright questions because authors generally receive no payment when companies buy second-hand copies and use them for training. US courts have already started addressing whether this type of use qualifies as transformative, while copyright rules in other countries can take a stricter approach.
The AirTag investigation now gives booksellers a clearer picture of where at least some of these unusual bulk orders are going, and it adds more pressure on AI companies to explain how they obtain copyrighted material for training.
Discussion