- Amazon reportedly buys large quantities of books for AI training and destroys originals during scanning
- A rare book was tracked via AirTag to an Amazon Las Vegas facility named LAS8 housing VGT3 operation
- Workers cut book bindings before scanning to speed up processing but permanently damage physical copies
Amazon is buying large quantities of printed books, scanning their contents for artificial intelligence training and destroying the physical copies in the process, according to an investigation by 404 Media.
The report by the publication said that it tracked a rare book using an Apple AirTag after a bookseller received an order for around 1,000 books through Biblio, a marketplace for used and rare books. The tracker followed the shipment across the US before the book reached an Amazon warehouse in Las Vegas.
404 Media reported that the shipment ultimately arrived at LAS8, an Amazon facility that also houses an operation identified as VGT3. Employees at the site reportedly said their work involves receiving, cutting and scanning books.
ALSO READ: Amazon's New User Agreement Seeks To Curb Class-Action Suits
The investigation said workers cut the bindings off books before scanning them, allowing the pages to be processed more quickly but destroying the original printed copy.
The reported activity comes amid growing demand from AI companies for large volumes of high-quality human-written material to train large language models.
Printed books are particularly valuable because much of their content is not readily available online. Books published before the rise of generative AI also provide human-created material that can help AI developers avoid problems associated with training models on AI-generated content.
404 Media said booksellers had reported an unusual increase in bulk purchases over the past year. Some sellers suspected AI companies were behind the orders because buyers appeared willing to pay for large quantities of seemingly unrelated books and showed little price sensitivity.
The investigation also linked the practice to Anthropic's Project Panama, which came to light during a lawsuit brought by authors against the AI company. Anthropic had acquired books from commercial marketplaces, cut their spines and scanned them for training data. A judge subsequently ruled that the scanning constituted fair use.
Amazon told 404 Media: “Amazon purchases books through commercial channels to help develop and improve the products and services our customers use.”
ALSO READ: Anthropic Eyes $6 Billion Decart AI Acquisition In Its Largest Known Deal
However, the company did not answer questions about why books are cut during the scanning process, how many facilities conduct similar operations or the backlash over the destruction of physical books.
Amazon is developing its own large language models under the Nova brand, which the company describes as frontier models capable of competing with systems developed by Google, OpenAI and Anthropic.
The books tracked by 404 Media were described as rare, including titles with limited circulation and books in less widely spoken languages. The investigation said VGT3 remains operational, even as some employees had previously expressed concerns about declining book shipments.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
