We Tracked a Shipment of Rare Books. It Ended at an Amazon AI Training Facility
404 Media tracked a shipment of ~1,000 rare books ordered through the Biblio marketplace using an Apple AirTag hidden inside one volume The books were delivered to Amazon's LAS8 facility in Las Vegas, specifically the VGT3 corner, which features a dinosaur-with-book logo at its entrance Internal Amazon worker forum discussions confirm that VGT3 destructively scans large volumes of books for AI training purposes This investigation corroborates earlier reporting (June 2025) about Anthropic's book
Analysis
TL;DR
- 404 Media tracked a shipment of ~1,000 rare books ordered through the Biblio marketplace using an Apple AirTag hidden inside one volume
- The books were delivered to Amazon's LAS8 facility in Las Vegas, specifically the VGT3 corner, which features a dinosaur-with-book logo at its entrance
- Internal Amazon worker forum discussions confirm that VGT3 destructively scans large volumes of books for AI training purposes
- This investigation corroborates earlier reporting (June 2025) about Anthropic's book scanning activities, suggesting multiple AI companies are acquiring physical books at scale
- The use of price-insensitive anonymous buyers on secondhand book marketplaces has become a widely suspected pattern for AI data procurement
Why It Matters
This investigation provides concrete, traceable evidence that Amazon operates a dedicated facility for destructively scanning books at scale, directly linking the physical book trade to AI training data pipelines. For AI practitioners and researchers, it underscores the growing reliance on proprietary and rare physical texts as training corpora, raising significant questions about copyright, licensing, and the sustainability of unlicensed data acquisition at industrial scale.
Technical Details
- Tracking methodology: 404 Media provided an Apple AirTag to a bookseller who received a ~1,000-book order on Biblio, enabling GPS-level tracking of the shipment's final destination
- Facility identification: The shipment was delivered to the VGT3 section of Amazon's LAS8 fulfillment center in northeast Las Vegas, identifiable by a dinosaur-with-book logo at the entrance
- Destructive scanning: Internal Amazon worker forum discussions confirm that VGT3 performs destructive scanning of large book volumes, meaning books are likely flattened or damaged during high-throughput digitization
- Data procurement pattern: The order was placed through a secondhand marketplace (Biblio) by an apparently anonymous, price-insensitive buyer—a pattern consistent with bulk AI training data acquisition strategies
Industry Insight
- AI companies are increasingly turning to physical book acquisition as a data source, bypassing digital licensing constraints but exposing themselves to legal and reputational risk around copyright infringement
- The destructive scanning model suggests that once books enter these pipelines, they are consumed as data rather than preserved, raising ethical concerns about the loss of rare and irreplaceable physical texts
- Regulatory scrutiny may follow as investigative journalism demonstrates traceable supply chains from book dealers to AI training facilities, potentially prompting legislation around data provenance and licensing for AI training corpora
Disclaimer: The above content is generated by AI and is for reference only.