Amazon Scans and Destroys Millions of Rare Books for AI Training Data
Updated
Updated · Forbes · Aug 19
Amazon Scans and Destroys Millions of Rare Books for AI Training Data
3 articles · Updated · Forbes · Aug 19
Summary
Millions of rare, antique and out-of-print books were bought by Amazon, scanned for AI training and then destroyed, according to a 404 Media investigation.
An Apple AirTag hidden in one ordered book helped trace shipments to VGT3, a secretive Amazon-owned Nevada warehouse where workers reportedly cut off bindings to digitize pages.
The practice reflects AI labs' scramble for fresh training material after largely exhausting the public internet by late 2024, pushing them toward scarcer copyrighted and physical sources.
VGT3's reported setup underscored the operation's scale and secrecy: the facility used a codename and displayed a logo showing a Tyrannosaurus rex about to eat an open book.