archive
· today in ai · 2026-08-17
Amazon buys rare books to train AI models
Archive item — written before sources were shown.
404 Media traced a tracker inside a purchased rare book to an Amazon facility in Las Vegas, and Amazon confirmed it scans out-of-print texts to train AI.
404 Media planted a tracker inside a rare book, bought it, and traced it to Amazon’s VGT3 facility in Las Vegas, where staff remove book spines to scan pages. Amazon confirmed the practice:
Purchases books through commercial channels to improve the products and services customers use.
It’s chasing never-digitized texts because retraining on AI-generated internet content risks “model collapse,” a documented quality drop.
sources
- 01Amazon, which started off selling books, is destroying rare texts to train AItechcrunch.com · reporting on the 404 Media investigation
Rami Steitieh