Skip to content
worth noting Security

ISBNdb withdrew an offer to sell printed books for AI training after a report by 404 Media

only one source so far

On 28 July, ISBNdb removed an offer to sell printed books for AI training from its website, nine days after a report by 404 Media. According to the company, it was merely a “test of market interest” – it had never bought or sold any books.

ISBNdb, a company operating a database of books and metadata, removed a page offering a service called “Printed Books Sourcing for Your AI LLMs Dataset Needs” from its website on 28 July – an offer to source printed books for companies training AI models. The move came nine days after a report by 404 Media, which drew attention to the offer around 21 July. On 30 July, ISBNdb added a statement about the matter to its homepage and news page.

According to the company, ISBNdb has never bought, scanned or sold any book – “neither for AI training nor for anything else”. According to its statement, the removed page was a “test of market interest”, and no such service was ever launched. The main service offered by the company, an API providing book metadata, continues to operate without restrictions, according to ISBNdb.

On the pages that have now been removed, the company previously claimed that printed books published before 2022 were ideal for AI training because they contained no AI-generated text, and that processing them could prevent so-called model collapse. The materials also suggested that authors who objected to this use of their work could write in ways that would sabotage AI models. The company further argued that buying used books on the secondary market did not deprive authors of income because these books had already fulfilled their financial obligation to their creators, and the marketing materials acknowledged that “the optics problem is real” and that a headline such as “AI company destroyed two million books” would not elicit sympathy.

According to its reporting, 404 Media spoke to booksellers who reported a recent increase in sales with unusual characteristics – orders with no apparent logic behind the choice of titles and overpaying for books regardless of their price, which one seller described as a typical sign of buying by AI companies.

What changed

Why it matters

The case shows how sensitive the issue of sourcing data for AI training on the secondary book market is – the company initially publicly advertised the sale of printed books to AI companies and, after media coverage, withdrew the claims and described them as merely a test of interest, without confirming who, if anyone, was behind the offer or who was buying the books. For booksellers and authors, the reported unusual increase in sales with suspicious characteristics is relevant; according to 404 Media, it may be linked to bulk purchases of books for AI training.

Relevant practical impact

What this means

01

For a business

The case illustrates the reputational risk for companies that publicly offer data sourcing for AI training without a verified business operation behind it – after media coverage, ISBNdb withdrew the offer and retrospectively described it as merely a “test of market interest”.

Risks and compliance
What to decide Companies offering or buying training data for AI should check whether public marketing claims about data sources reflect reality and consider the reputational impact before publishing similar offers.
More business impacts →
AI training data copyright book authors copyright ISBNdb printed books

Check the original

Event sources

only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.

1
404 Media (investigative tech journalism) independent context · first detected Company Offering Printed Books to Train AI Stops After 404 Media Coverage