Key Points
AirTag tracked rare book from California to Amazon's Las Vegas warehouse VGT3.
Amazon employees cut book spines and scan pages for data.
AI companies need rare books because internet text contains AI-generated content.
Amazon declined to confirm AI training purpose but rivals claim they avoid rare books.
A 404 Media investigation has confirmed that Amazon is buying rare books in bulk, scanning them for AI training data, and destroying them in the process. Reporter Emanuel Maiberg placed an Apple AirTag in a rare book as part of a 1,000-book order and tracked it from California to an Amazon warehouse in Las Vegas known as LAS8. The facility houses a team called VGT3 that cuts book spines and scans pages. Amazon declined to confirm the AI training purpose but said it purchases books to improve customer products.
How the AirTag revealed Amazon’s operation
404 Media partnered with a bookseller who agreed to plant an AirTag in a book from a bulk order placed through Biblio. The book travelled from a California airport to a distribution warehouse near Kenosha, Wisconsin, then by truck through Grand Junction, Colorado, before arriving at Amazon’s Las Vegas facility. The AirTag placed the book at the north end of LAS8, in a section Amazon employees identified as VGT3, a separate operation with its own entrance and logo: a Tyrannosaurus rex holding an open book.
Why AI companies need rare books for training
Large language models learn by processing text collections, so training data quality matters significantly. AI companies prefer rare and antique books because they contain structured information that has already passed quality control, unlike internet text contaminated by AI-generated content. Booksellers across Australia, Europe, and the United Kingdom have reported a surge in anonymous bulk orders for obscure titles over the past year. Court documents have shown this is a common practice, with companies like Anthropic reportedly destroying millions of books.
Amazon’s denial and rival claims
Amazon declined to comment on specific findings but provided a statement saying it purchases books through commercial channels to develop and improve customer products. The company did not mention AI training. However, rivals like Anthropic and xAI have publicly stated they do not train on rare or antique books. Amazon employees at VGT3 told workers’ forums that the company had run low on books to scan earlier this year, suggesting urgent demand for training material.
What this means for investors and the AI race
The discovery highlights the hidden costs of the AI boom. Companies competing for frontier models need massive amounts of unique training data, creating pressure to acquire rare materials by any means. With Apple’s AirTag technology enabling this kind of tracking, supply chain transparency is becoming harder to avoid. For investors in AI companies and their suppliers, the ethical and legal risks around data sourcing are rising as scrutiny intensifies.
Final Thoughts
Amazon’s book-destruction operation confirms what booksellers have suspected for months: AI companies are systematically acquiring and destroying rare books for training data. The AirTag investigation provides concrete proof of a practice that competitors claim they avoid, raising questions about competitive advantage and regulatory risk in the AI sector.
FAQs
A reporter placed an Apple AirTag in a rare book from a bulk order and tracked it to Amazon’s Las Vegas facility, where employees confirmed they cut and scan books.
Rare books contain high-quality, structured information that has passed editorial review, unlike internet text polluted with AI-generated content.
No. Amazon said it purchases books to improve customer products but did not mention AI training specifically.
The article does not specify a total number. Employees reported the facility had run low on books to scan earlier in 2026.
Disclaimer:
The content shared by Meyka AI PTY LTD is solely for research and informational purposes. Meyka is not a financial advisory service, and the information provided should not be considered investment or trading advice.
About Author

Danny Kontos
Co FounderDanny Kontos has been a stock investor since 2007 and co-founded Meyka in 2023. He keeps a small, focused portfolio and only moves when the numbers are hard to argue with. He has waited years on a single position before. Before Meyka, he ran a web hosting company and a mortgage lending platform, so he knows what a well-run business actually looks like under the hood. This article did not come from a news cycle. It came from someone who has been watching this space for a long time.
What brings you to Meyka?
Pick what interests you most and we will get you started.
I'm here to read news
Find more articles like this one
I'm here to research stocks
Ask Meyka Analyst about any stock
I'm here to track my Portfolio
Get daily updates and alerts (coming March 2026)