AI Companies Are Buying Millions of Books, Scanning Them and Destroying the Copies

A France 24 Focus investigation broadcast on 16 October 2026 exposed a practice by which artificial-intelligence companies are bulk-buying physical books from the secondhand market through anonymous middlemen, scanning the texts to create training datasets for large language models, and then destroying the physical copies. One Dutch bookseller described receiving what he initially dismissed as a spam order for 3,000 identical copies of the same title before realising it was genuine. The scale of purchasing has created unusual spikes in secondhand and antiquarian book markets, and conservationists and librarians warn that rare volumes — including books with only a handful of surviving copies — risk disappearing permanently from the public record. Authors and publishers argue the practice exploits a legal grey zone: because companies buy the physical object, they argue no copyright licence is needed, yet the creators of the text receive no compensation for its use in AI training. The French journalist team's report also highlighted the broader cultural stakes, interviewing book collectors and heritage professionals who described the systematic acquisition and destruction of irreplaceable printed artefacts as a quiet cultural crisis with no regulatory oversight. The story spread rapidly across YouTube and international news aggregators in the days following publication.