Amazon buys and destroys rare books to train AI models
Original: Amazon, which started off selling books, is destroying rare texts to train AI
Why This Matters
Highlights escalating demand for pre-AI human text and raises concerns over destruction of irreplaceable cultural materials.
Amazon is purchasing rare books, removing their spines, and scanning them as AI training data, according to 404 Media. A tracking device placed inside one rare book led investigators to Amazon's VGT3 facility in Las Vegas. Amazon confirmed it 'purchases books through commercial channels to improve products and services.'
Investigative outlet 404 Media reported that Amazon is systematically acquiring rare and out-of-print books, physically destroying them by cutting off their spines, and scanning the contents to use as training data for its large language models (LLMs). The operation was confirmed when 404 Media placed a tracking device inside a rare book, which was ultimately traced to an Amazon facility called VGT3 in Las Vegas — identified by a logo depicting a dinosaur holding a book. Amazon responded with a statement saying it 'purchases books through commercial channels to improve the products and services customers use,' without elaborating further. Rare and out-of-print books are considered especially valuable training data because they are not accessible on the internet and, crucially, predate the era of LLM-generated text. AI-generated text poses a risk known as 'model collapse,' where a model's output quality degrades after ingesting too much synthetic content. The move follows similar controversies, including Anthropic's use of allegedly pirated books for training data.