AI Companies Accused of Buying and Destroying Books for Training Data

Major technology companies are acquiring massive quantities of printed books to scan for AI training data and destroying the physical copies in the process. A 404 Media investigation tracked a shipment of rare books using an AirTag, which revealed the books ended up at an Amazon warehouse in Las Vegas, Nevada. Employees at that facility reported receiving large shipments of books and cutting off the bindings to facilitate faster scanning. Amazon stated that it purchases books through commercial channels to improve its products and services.

Second hand booksellers in the United Kingdom and Ireland have reported a surge in strange bulk orders from buyers in the United States, Canada, and Europe. These orders often lack editorial coherence, mixing unrelated topics such as 18th century African agricultural tools and 1950s racing driver biographies. Some of these requests came from Zoom Books, a Canadian company that specializes in scanning books and destroying them. Zoom Books has denied involvement in AI training but has not disclosed its clients.

Anthropic, the creator of the AI model Claude, previously engaged in a similar operation known as Project Panama. The company invested millions of dollars to acquire and scan books to improve the writing quality of its model and avoid low quality internet language. Anthropic agreed to pay approximately 1.3 billion euros to settle a class action lawsuit brought by authors over copyright concerns.

Center leaning coverage focuses on the disruption to the second hand book market and the legal battles over copyright. Center left leaning coverage emphasizes the investigative tracking of Amazon's operations and the technical motivation for using printed books to avoid model collapse caused by AI generated text.