Investigation Finds Rare Books Dismantled and Scanned to Train AI Systems

An investigation by journalist Emanuel Maiberg, co-founder of the outlet 404 Media, has uncovered what he describes as a globalised and deliberately opaque supply chain for acquiring and destroying books to create training data for artificial intelligence systems, according to an interview aired on France 24's Spotlight programme.

Maiberg, speaking with host François Picard, said his reporting traced how books are purchased through anonymous marketplaces before being sent to large-scale scanning facilities. At these facilities, he said, the bindings of the books are removed so that individual pages can be processed at high speed by scanning equipment.

According to Maiberg, the scale of the practice is difficult to quantify precisely but could involve thousands, if not millions, of books being fed into what he characterised as an industrial process geared toward generating data for AI models.

Maiberg told France 24 that he does not believe the destruction of the books stems from malicious intent on the part of those involved in the supply chain. Rather, he described the driving force as indifference toward the physical books themselves, with the focus placed solely on extracting the text they contain for use in training AI systems.

The investigation, as presented in the interview, points to a lack of transparency at multiple stages of the process, from the anonymous marketplaces where books are initially bought to the scanning facilities where they are ultimately dismantled. Maiberg's reporting suggests that rare books have been swept up in this process alongside more common titles, though the interview segment did not specify further details on the identities of the companies or individuals involved in the supply chain.

France 24 presented the discussion as part of its Spotlight programme, which featured Maiberg discussing the findings of his investigation into the practice.