Amazon, once an online bookseller, is destroying rare books to train AI models

Amazon, the company that began as an online bookseller, is now buying large quantities of rare books, cutting off their spines and scanning them to train AI models, according to a TechCrunch report based on an investigation by 404 Media.

404 Media said it placed a tracking device inside a rare book and traced it to an Amazon facility in Las Vegas known as VGT3, which identifies itself with a symbol of a dinosaur holding a book in its claws.

In a statement, Amazon told 404 Media that it "purchases books through commercial channels to improve the products and services customers use."

The report explains the motivation: companies like Amazon need unfathomably large amounts of text to train their LLMs, and models have already ingested much of what is available on the internet — Anthropic, for example, was accused of illegally pirating books.

Rare books, especially ones that are out of print or impossible to find on the internet, offer a new source of coveted training data. These texts are especially valuable since there's no chance that anything published before 2022 was written by an LLM.

When LLMs train on AI-generated text, they risk "model collapse," a degradation in output quality — which makes human-written rare books an increasingly attractive data source for model builders.

The revelations sharpen the ongoing fight over training data and copyright, a battle already playing out in lawsuits from publishers and authors. What remains to be seen is how Amazon responds to the tracking-device findings, whether the book industry pushes back, and whether regulators take a closer look at large-scale scanning of physical books.