This Dutch bookseller thought a request for 3,000 copies was 'spam or phishing.' Instead, AI companies are scanning and destroying books to train AI

A Dutch bookseller recently received an unusual email from an entity called 2077AI. The request asked to source over 3,000 specific book titles with instructions to ship them to China. Pieter de Vries, the bookseller, immediately identified the message as potential spam. He did not act on the request, but soon discovered that other booksellers across the Netherlands had received identical inquiries.

This incident highlights a hidden supply chain currently feeding the artificial intelligence industry. Major AI firms are looking beyond the open internet to find high-quality text for training their models. Reports indicate that companies are purchasing large volumes of physical books to undergo a process known as destructive scanning. This method involves removing the binding of a book to feed pages through high-speed scanners before the physical copy is discarded.

Legal records from previous copyright cases confirm that this practice happens at scale. While some companies state they use commercially acquired data, the procurement of millions of physical volumes remains a clear priority for training advanced systems. Other services have even advertised their ability to source hundreds of thousands of books specifically for LLM training needs.

For the bookselling industry, these bulk requests present a strange development. Dealers of rare and antiquarian books are accustomed to individual sales, not high-volume orders intended for digitization. As artificial intelligence models continue to demand more data, the physical world of print serves as a growing target for these organizations.