Amazon’s AI book hunt is turning rare volumes into disposable data

Amazon’s AI book hunt is turning rare volumes into disposable data
The trail of a rare book has become a warning flare for the AI boom: a volume bought in bulk can now end up not on a reader’s shelf, but under a blade.

Booksellers had spent months watching unusual, high-volume orders for obscure titles and wondering whether AI companies were behind them. The suspicion sharpened when 404 Media placed a tracker inside a shipment of rare books and followed it to Amazon’s VGT3 facility in Las Vegas. Workers there said they receive large book shipments, cut off the bindings and scan the pages; “The printed book is destroyed in the process.”

The logic is straightforward, if unsettling. AI developers need vast supplies of distinctive text, and out-of-print or poorly digitized books offer material that is both hard to find online and less likely to contain AI-generated prose. Amazon is developing frontier models, while workers’ reported barcode and ISBN checks have reinforced booksellers’ concern that companies are methodically hunting for titles missing from their datasets.

Amazon has not explicitly described the Las Vegas operation as AI training. Its public response is broader: it “purchases books through commercial channels to help develop and improve the products and services our customers use.” That framing clashes with the preservation argument. Sellers say an inexpensive, little-known book can still carry historical, scholarly or sentimental value—and may be the last surviving copy of an edition.

The dispute is not over whether old books can be digitized, but how. The Internet Archive says it avoids severing bindings, relying on careful manual handling because automated systems can damage brittle and special-collection volumes. “Clean, dry human hands are the best way to turn pages,” its scanning operation has said.

Some observers note that not every dusty title can be saved forever, and bulk sales can help independent sellers. But the tracker’s route has made the stakes concrete: faster, cheaper AI data acquisition may be colliding with the slower work of keeping cultural records alive.

Continue reading https://foxvector.com/stories/01a01154-3ecb-2c81-7180-33603f9bab56