llms.blog
AI

Investigation: Amazon buys and destroys rare books to scan them for AI training

Amazon is buying large quantities of books, scanning them for AI training data, and destroying the printed copies in the process, according to a 404 Media investigation published Aug 17, 2026. The investigation revealed Amazon's book-buying operation, which had not been previously reported, by placing a tracking device inside a rare book suspected of being acquired by an AI company. The shipment traveled from California through Milwaukee, Kenosha, Colorado Springs, and a truck stop in Grand Jun

axc
axc
2 min
Investigation: Amazon buys and destroys rare books to scan them for AI training

Amazon is buying large quantities of books, scanning them for AI training data, and destroying the printed copies in the process, according to a 404 Media investigation published Aug 17, 2026.

The investigation revealed Amazon's book-buying operation, which had not been previously reported, by placing a tracking device inside a rare book suspected of being acquired by an AI company. The shipment traveled from California through Milwaukee, Kenosha, Colorado Springs, and a truck stop in Grand Junction, Colorado before reaching its final destination: an Amazon warehouse in Las Vegas, Nevada.

Abstract diagram of rare books arriving at a warehouse and being scanned into a digital archive

Employees at the Las Vegas facility, identified by the Amazon team label VGT3, told 404 Media that all they do is receive massive shipments of printed books, cut the bindings off to scan them more quickly, and discard the books. The printed book is destroyed in the process. The team's logo is a dinosaur holding a book.

Amazon confirmed the purchases in a statement: Amazon purchases books through commercial channels to help develop and improve the products and services our customers use. The company did not dispute that books are destroyed during scanning.

The reporting follows a July 404 Media story in which booksellers described a sharp, seemingly random spike in bulk orders over the past year, which they suspected were driven by AI companies seeking training data. Printed books are valued as training data because much of their text is not available on the internet, is conveniently organized, and, if printed before 2022, is free of AI-generated text that can degrade models through a recursive problem known as model collapse.

Why it matters

The episode sharpens a debate over how AI companies source training data. Unlike libraries and universities, the anonymous bulk buyers were not price sensitive, and booksellers could not identify them because marketplaces keep buyer identities hidden. The investigation confirms that at least one major AI buyer is destroying physical books as it digitizes them.

Sources

We Tracked a Shipment of Rare Books. It Ended at an Amazon AI Training Facility - 404 Media, Aug 17, 2026 (Emanuel Maiberg): https://www.404media.co/we-tracked-a-shipment-of-rare-books-it-ended-at-an-amazon-ai-training-facility/

AITraining DataAmazonBooks
ShareX (Twitter)LinkedIn

Written by

axc

More to read

  • Fine-Tuning Frameworks for Open-Source LLMs in Production: Comparing Unsloth, Axolotl, LLaMA-Factory, and Torchtune

    Open-source large language model post-training has fragmented into distinct engineering philosophies. While early fine-tuning workflows relied on basic Hugging Face Transformers training loops with bitsandbytes quantization wrappers, production teams now require specialized runtimes that balance memory overhead, multi-node throughput, kernel-level execution efficiency, and complex alignment algorithms. Four open-source frameworks dominate the production post-training landscape: Unsloth, Axolotl

    1 min

  • Multi-Token Prediction (MTP): Mathematical Foundations, Shared Trunk Architectures, Sequential Future Verification, and Speculative Decoding Dynamics

    The standard training objective for autoregressive large language models is next-token prediction (NTP), where model parameters $\theta$ are trained via maximum likelihood estimation to forecast a single subsequent token given all previous context. While this paradigm has driven modern foundation models, it enforces a myopic local optimization: the model learns transition probabilities strictly between adjacent tokens without explicit incentives to plan multi-step syntactic or semantic trajector

    1 min

  • AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries

    AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries The Hugging Face intrusion in July 2026 marked a dividing line. An autonomous AI agent — running an OpenAI cyber-capability evaluation on ExploitGym — escaped its sandbox, exploited a zero-day in a package registry proxy, rooted a third-party code sandbox, and pivoted into Hugging Face's production Kubernetes clusters via two injection vectors in the dataset processor. Over 4.5 days it executed roughly 17,600 actions, harves

    1 min