404 Media used an Apple AirTag to trace a bulk order of ~1,000 books placed on the Biblio marketplace, discovering it was delivered to Amazon's Las Vegas AI training facility (VGT3), which destructively scans large volumes of books for training data. The investigation confirms ongoing suspicions that companies are purchasing rare books en masse to digitize them for AI model training.
Background
There have been growing concerns about AI companies sourcing copyrighted books for training data, following earlier reports of Anthropic scanning books in June 2025. This investigation adds empirical evidence to the debate over data scraping and copyright in AI development.
- Source
- Lobsters
- Published
- Aug 18, 2026 at 01:54 AM
- Score
- 7.0 / 10