Amazon is destroying rare books to train AI models

Major tech companies continue their aggressive pursuit of massive text datasets to train large language models, especially as readily available internet data has been thoroughly ingested. Recent investigations reveal that e-commerce giant Amazon has been acquiring large quantities of rare books, slicing off their bindings, and scanning the pages to convert them into training data for artificial intelligence systems.

This operation came to light after reporters at 404 Media placed a hidden tracking device inside a rare book, which ultimately led them straight to an Amazon operational facility located in Las Vegas, Nevada.

Stock photo for illustration only, not from the actual event

The facility, designated as VGT3, features a distinct symbol depicting a dinosaur holding a book in its claws. This strategy highlights how tech firms are now targeting out-of-print and offline literature as a fresh, coveted source of training data that cannot be found anywhere else on the internet.