Suspicions that rare books are being purchased in bulk, scanned, and then destroyed for the purpose of training artificial intelligence models have been confirmed by a new development. An AirTag placed inside a small book by a bookseller proved that the book went to Amazon’s AI training facility in Las Vegas.

It was determined that at this facility, the books were separated from their bindings, their pages scanned, and then destroyed. It was even noted that the logo at the entrance of the facility featured a T. rex figure eating a book.

Amazon’s Statement and Process

In its statement regarding the matter, Amazon stated that it purchases the books through commercial channels and uses them to improve its work. The company avoided directly addressing the issue of AI training.

It is known that Amazon largely needs original training data to remain competitive. In previous interviews with company employees, it was stated that the stock of books to be scanned had run out and the facility was facing the risk of closure.

It is now confirmed that the facility continues its operations and that rare books received are systematically destroyed. While employees state that this is preferred due to the flexible hours, book lovers are reacting negatively to this destructive scanning system.

Methodical scanning via ISBN

Research shows that AI companies aim to systematically scan all printed books in the world using ISBN numbers. The fact that Amazon employees check barcodes or ISBNs before scanning books reinforces this theory.

Booksellers argue that this creates an ethical dilemma and that the pricing process is being bypassed. AI companies are focusing on lesser-known or untranslated older books rather than high-value first editions.

Although these books may have low monetary value, they can hold great historical and intellectual significance. Booksellers say that companies view the content of these works only as a data dump.

The debate continues.

On platforms like Reddit, users are debating how problematic it is to get rid of books gathering dust on shelves. Some users argue that these books are not actually read, while others worry that historical information will remain locked away in the hands of AI companies.

Critics argue that AI models will only offer censored or shared modules from these books. Some booksellers emphasize that works without academic value can be sacrificed, but rare copies should be preserved.

Companies’ efforts to keep educational data anonymous make it difficult to achieve transparency in this matter. Do you think it’s acceptable to destroy rare books for AI training?