Sunday, August 23, 2026

Canadian Firm Seeks 800,000 Dutch Books for AI Training?

Valyrian News Network 4 min read

Canadian Firm Seeks 800,000 Dutch Books for AI Training?

De Slegte, the prominent Dutch-language bookstore chain active in Belgium and the Netherlands, has received an unexpected email from Canada expressing interest in purchasing 800,000 Dutch-language books. The bulk acquisition request from Zoom Books, a Canadian secondhand book dealer, has raised questions about the purpose and destination of the books, with many in the industry suspecting they may be destined for AI training programs.

The Email and De Slegte’s Response

“At this moment we have a wish list of 800,000 Dutch-language books,” wrote Manroop Gill, one of the co-founders of Zoom Books, in an email addressed to De Slegte. The Canadian company asked the bookstore chain to send its inventory with ISBN codes so they could select which books to order, according to VRT NWS.

De Slegte declined to respond to the email. “Everyone can see on our website what we offer for sale and order what they want,” said Kris Landuyt, spokesperson for De Slegte. The chain, a major player in the remaindered and secondhand book market, reportedly received requests spanning “ordinary novels to books with huge print runs,” according to AD.

A Global Pattern of Bulk Purchases

Zoom Books has previously bought large quantities of books in Germany and Spain. According to Swiss broadcaster SRF, the company purchased pallets of books from a German online bookstore—specifically non-fiction titles from 1970 onwards with ISBN numbers, described as “dusty warehouse leftovers that nobody wanted for years.” Only one copy of each title was purchased.

Spanish newspaper elDiario.es reported that Zoom Books previously had five blog posts about buying secondhand books to feed AI algorithms and then destroying them. Those posts have since been removed, though the company’s website still has an empty page titled ‘ai-training-books.’

Booksellers in Spain described the purchases as a form of “literary plundering.” “We are not only merchants but have functions of preservation, conservation and restoration of bibliographic heritage,” said Miguel Ángel Ortega, president of UNILIBER, the Professional Association of Antique Book and Collecting.

The AI Connection

The pattern of purchases matches what has been observed globally—AI companies buying physical books to scan and destroy them for training data, free of AI-generated content. AI company Anthropic was previously revealed to be mass-purchasing books through secondhand dealers to train its AI models using “destructive scanning”—cutting off spines, scanning pages, and discarding the remains.

AI companies face what experts call the “data wall”: without new, unpublished, and long-form texts to train algorithms, AI risks cognitive stagnation and could end up feeding on documents created by its own technology. Books printed before the internet era are valuable because they are guaranteed to be free of AI-generated content.

404 Media reported that AI companies are buying tons of old books because they’re free of AI slop, with intermediaries like ISBNdb offering bulk book acquisition services.

Denials and Silence

Zoom Books denies any direct connection to AI companies. “Zoom Books is a secondhand bookstore. We buy surplus and used stock and resell it,” the company stated to elDiario.es. A spokesperson told The Guardian: “We acquire secondhand books and resell them intact. We do not digitize books. Our priority is always reuse, and when a book can no longer be rehomed, it is responsibly recycled.”

However, the company did not disclose whether it sells intact books to AI companies, saying: “Our commercial agreements require confidentiality, so we do not disclose or discuss our customers.” VRT NWS asked Zoom Books why it wants such a large quantity of Dutch-language books and whether they might be sold to AI companies, but received no response.

Anthropic, for its part, stated: “None of our data acquisition programs buy and destroy rare or antiquarian books.”

A US judge ruled in June 2025 that destructive scanning falls under ‘fair use’ under US copyright law. Anthropic settled a lawsuit for potentially $1.5 billion for books it downloaded online without author permission.

For the Dutch-speaking world, the stakes are particularly high. Dutch is a relatively small language, and if 800,000 Dutch-language books are removed from circulation and destroyed, it could have a significant impact on Dutch-language cultural heritage. The Next Web noted that booksellers fear their rarest books are being pulped to feed AI, with the trail leading to Zoom Books.

What to Watch For

As the AI industry continues to search for clean training data, the pressure on secondhand book markets worldwide is likely to intensify. Booksellers, cultural institutions, and policymakers will need to grapple with the tension between AI development and the preservation of irreplaceable cultural artifacts. For now, De Slegte’s decision to ignore the email from Zoom Books represents a small but notable act of resistance in an increasingly complex global debate.