According to German booksellers, demand has surged in recent months, DW reports. On industry forums, sellers say that some individual orders are comparable in size to their annual revenue. In some cases, books that had remained unsold for years have found buyers. Orders typically come through the Canadian-American company Zoom Books.
Scientific and academic literature is in the highest demand, along with books on regional history, linguistics, economics, and law. As noted by the Swiss broadcaster SRF, AI developers are interested in texts with historical linguistic features and stylistic elements that have not yet been digitized.
Interest in paper books has intensified after developers of the largest language models have virtually exhausted the volume of publicly available texts on the internet. The Washington Post previously reported that the American company Anthropic, as part of the Panama project, purchased millions of printed books to digitize them for training artificial intelligence. To speed up the process, the book covers were cut off, effectively destroying the physical copies.
German bookstore owners fear that a similar practice will spread to Europe. Berlin bookseller Urban Zerfas compared what is happening to a gold rush. According to him, sellers do not want valuable editions to be destroyed after purchase, and he called on the German Booksellers’ Association and the authorities to address the issue. He believes that such a practice may violate copyright laws.
The issue remains controversial from a legal standpoint as well. According to SRF, American companies can invoke the principle of “fair use”, which in certain cases permits the copying of copyrighted works without the rights holder’s consent. It is precisely this mechanism that is now becoming one of the key arguments in legal proceedings surrounding the training of artificial intelligence systems.
There are also concerns about the potential loss of cultural heritage. Roger Sonnewald, an antiquarian bookseller from Tübingen, told Südwestrundfunk that the most frequently purchased items are publications from the 1970s and later that are still protected by copyright. According to him, earlier German-language literature has largely already been digitized and is accessible through libraries, whereas books from the second half of the 20th century are of the greatest value to AI developers.
Follow our updates
