AI's book hunt sparks secret destruction fears

Artificial intelligence companies are increasingly turning to anonymous middlemen to buy vast numbers of physical books, according to new reporting, fuelling fresh concerns over transparency, copyright and the fate of rare printed works.

The practice highlights the growing demand for high-quality human-written material as technology firms race to build more powerful AI systems. Critics say millions of books are being scanned and discarded behind the scenes, while companies seek to avoid public backlash.

Behind The Anonymous Buying Spree

According to 404 Media, AI developers have been using third-party firms to source books in bulk rather than purchasing them directly. The approach allows companies to obtain large collections while keeping their identities hidden.

One company, ISBNdb, previously advertised services aimed at supplying books "tailored to your LLM training needs" at industrial scale. Archived versions of its website also referred to strict non-disclosure agreements to protect clients' identities, although the company has since removed those pages and says it has moved away from that business direction.

Why Printed Books Suddenly Matter

As AI-generated content floods the internet, older printed books have become increasingly valuable as sources of original, human-authored writing. Rare and out-of-print titles are particularly sought after because they may contain material unavailable elsewhere.

The issue gained prominence after a 2025 court ruling revealed that Anthropic had purchased, dismantled, scanned and discarded millions of books while training its Claude AI models. The court found the scanning process could qualify as fair use under US copyright law, although internal documents reportedly showed company officials recognised the reputational risks.

Rare Books Vanish As Demand Surges

Booksellers say demand has accelerated sharply in recent months. One dealer told 404 Media he is now shipping hundreds of books each week, compared with around 20 previously, with many orders focusing on obscure or low-circulation titles.

Dutch booksellers have reported similar trends, with buyers targeting books identified through ISBN databases. While some dealers welcome the commercial boost, others say they are uneasy about uncommon works being destroyed after digitisation.

The controversy adds to mounting scrutiny of the AI industry as legal battles over copyright, data collection and transparency continue to intensify, with publishers, authors and technology firms increasingly at odds over how the next generation of AI models should be trained.

AI companies are anonymously buying and destroying millions of books through middleman services