Japanese used bookstores are receiving unusual orders. Starting last month, some stores have begun selling hundreds of books daily, while others have recorded periods where sales exceeded the norm by five times. There is suspicion that some of these orders are intended for digitization to train artificial intelligence systems, after which the copies are destroyed.
According to Japanese broadcaster NTV, about 50 tons of Japanese books have been sent to the United States since last year. What is unusual is not only the volume but also the choice of buyers: they avoid the most popular publications and order works on Japanese philosophy, history, medicine, law, politics, and culture. Orders for books on the Edo period and old political publications are also coming in.
NTV discovered hundreds of orders placed under different accounts but with the same address at a logistics center in Okayama city. From there, the works were sent to their final destinations. The company managing this center declined to disclose the identity of the buyers to the Japanese press.
It is assumed that part of these purchases is related to training artificial intelligence systems. Companies in this industry are already purchasing books in large batches for content digitization and use in training large language models (LLMs). In most cases, the physical copies are disassembled, pages are scanned, and the text is converted into digital files, after which the books can be disposed of.
This practice drew attention in the US during a case involving Anthropic. Documents from this case showed that the company organized an operation to purchase millions of physical books, digitize the content, and subsequently destroy the copies. The goal was to form a digital fund for the development of Claude. The company was fined $1.5 billion (about 7.6 billion Brazilian reais) for copyright infringement.
In another, more recent case, the portal 404 Media tracked a rare book that was sold in a batch of a thousand copies. The material arrived at an Amazon facility in Las Vegas, which, according to the site, was used for large-scale book digitization and destruction. The operation was allegedly conducted by the company VGT3, located in the retailer's logistics complex, to convert physical works into digital text before their disposal. Amazon confirmed only that it purchases books through commercial channels for 'developing and improving products and services used by customers,' but did not detail the use of the material. However, the logo of the division appears to refer to this process.