A similar wave hit Vienna bookseller Karin Schaden — orders came in at two in the morning via a book platform. "It looked as if they weren't being placed by an actual person. I think they were automated bots," she told Die Presse. By her estimate, the Vienna antiquarian bookshop alone could have earned several tens of thousands of euros. The bulk purchases took place from late April to early June and have since stopped; they were also reported in Switzerland and Spain. Die Presse ↗
Suspicion: fodder for artificial intelligence
Why would anyone buy up unsellable books in bulk? AI developers need enormous quantities of text for training, and according to both Tagesschau and Die Presse, digitally available books have already largely been processed — so AI companies are looking for additional sources of text. Old printed titles are thus not only a new source of text but, according to Die Presse, also a possible legal workaround. Die Presse ↗The Canadian company Zoom Books, through which the purchases were made, did not directly answer questions from the editorial teams, according to both Tagesschau and Die Presse. However, its manager Reed Pannell told the German newspaper taz that the company "does not digitize or destroy books, contrary to recent speculation." He would not reveal to whom or for what purpose it resells them. Die Presse ↗
"Project Panama": how Anthropic destroyed books
The booksellers' suspicion did not arise out of nowhere. In January, the Washington Post revealed Anthropic's previously secret "Project Panama," run by the creator of the chatbot Claude. The company hired a scanning service that, over the course of six months, cut apart an estimated half a million to two million books — a hydraulic cutter sliced off the spines, the pages went through high-speed scanners, and the remaining copies were sent for recycling. An internal 2024 document requested that "it not be known that we are working on this." Washington Post ↗In the authors' lawsuit, the court ultimately ruled that the use of legally purchased books for training itself constituted "fair use" — since the physical copies became the company's property upon purchase. At the same time, however, it found that before launching the project, Anthropic had violated copyright by downloading millions of pirated e-books. Last year, the company agreed to pay authors 1.5 billion dollars. Washington Post ↗
The battle over copyright
Germany's Börsenverein, the book trade association, considers this kind of training practice to be a clear violation of copyright. "Under German law, scanning books — for whatever purpose — would not be permissible," warned association spokesperson Thomas Koch. However, the law of the country where the scanning takes place applies, and American courts are more lenient on "fair use." Tagesschau ↗Disputes are mounting. According to Tagesschau, publisher Penguin Random House is suing OpenAI because its model, given a simple prompt, allegedly copied a children's book almost word for word. In this context, Tagesschau also writes about companies and intermediaries through which physical books are turned into digital copies ready for further use. Tagesschau ↗
What hotinfo is following
- Whether Zoom Books will reveal to whom it resells the books, or whether any AI company will confirm the identity of the end buyer.
- The verdict in Penguin Random House's lawsuit against OpenAI — which could become a precedent for training on copyrighted texts.
- The stance of the EU and German authorities: whether American "fair use" will hold up under stricter European copyright law.







