AI companies are anonymously buying and destroying millions of books through middleman services to avoid headlines about AI companies buying and destroying millions of books

AI developers are resorting to third-party middleman services to quietly purchase and destroy physical books in a bid to harvest high-quality training data for large language models (LLMs) without triggering public backlash.

By routing orders through intermediaries, AI firms obtain pristine, human-authored text that predates the proliferation of AI-generated content while attempting to avoid public relations damage. Physical books—particularly rare or low-circulation titles absent from competitors' datasets—have become prized training material. Once acquired, the books are typically de-spined, scanned into digital datasets, and discarded or pulped.

Intermediaries and Confidentiality

To keep these acquisitions discreet, AI developers have relied on commercial middleman platforms to aggregate and source inventory. PC Gamer reported that database provider ISBNdb previously offered specialized book-sourcing options "tailored to your LLM training needs, delivered at the scale AI demands."

To insulate tech clients, the service promised a "strict NDA on every engagement," guaranteeing that "your identity, strategy, and acquisition targets are never disclosed." In a blog post describing the service, ISBNdb acknowledged the sensitivity surrounding physical destruction, noting that "'AI company destroys two million books' is not a headline that generates sympathy."

Following public attention on the offerings, ISBNdb scrubbed the AI-specific pages from its website. In a news update, the company stated the service was "part of exploring demand, and we've chosen to pivot away from that direction."

Impact on Booksellers and Legal Precedent

Booksellers specializing in rare and low-circulation inventory have felt the impact directly since April 2026, reporting unprecedented surges in orders. These bulk purchases frequently consist of seemingly random titles tied together only by the fact that they have registered ISBNs.

The buying spree leaves dealers conflicted over selling off rare inventory destined for destruction. PC Gamer quoted one anonymous bookseller who detailed the mixed financial and ethical impact: "I personally have mixed feelings about all of this. It benefits me financially as well as by clearing out old inventory that is otherwise unlikely to sell. I've been well-suited for these sales with inventory from overseas and foreign language books. On the other hand, I don't like the end-use, and I don't like that uncommon books are being pulped."

While public relations remains a concern for tech firms, the legal framework has favored the practice. A 2025 court ruling established that Anthropic's practice of de-spining, scanning, and discarding millions of books for AI training constituted "fair use" under Section 107 of the Copyright Act. However, tech developers remained aware of the optics; an internal Anthropic planning document obtained by PC Gamer explicitly noted, "We don't want it to be known that we are working on this."

---

• PC Gamer Hero image: https://cdn.mos.cms.futurecdn.net/8hfqaZRx4qSq9BP2Nvuuob-1280-80.jpg