A book database quietly removed its AI training offer after media scrutiny. ISBNdb now says it never sold, scanned, or bought books for AI. The company claims the service was only a market test.
ISBNdb, a well-known book metadata provider, has withdrawn its offer to source printed books for AI training after facing criticism and media attention. The move comes just days after 404 Media reported that ISBNdb was marketing a service to supply physical books to AI companies seeking data for large language models.
Following the coverage, ISBNdb deleted the relevant landing pages and issued a statement on its homepage and news section. The company said it had never purchased, scanned, or sold any books for AI training or other purposes. According to ISBNdb, the now-removed service was only a test to gauge market interest and was never operational. The company emphasized that its core business remains providing book metadata to help bookstores, libraries, and reading apps connect readers with titles, not handling the books themselves.
The landing page, previously titled “Printed Books Sourcing for Your AI LLMs Dataset Needs,” was taken down on July 28. ISBNdb stated that the experiment was part of exploring demand but has since been abandoned. The company assured users that its main metadata API services remain unaffected and continue to operate as usual.
ISBNdb responded to 404 Media’s request for comment with the same statement now posted on its homepage. The company did not respond to earlier inquiries from the outlet.
Archived versions of ISBNdb’s site reveal that the company had promoted the idea that printed books published before 2022 were ideal for AI training, as they would be free from AI-generated content and could help prevent model collapse. The removed materials also suggested that authors concerned about their work being used for AI could attempt to manipulate or sabotage AI models through their writing.
Booksellers interviewed by 404 Media reported a recent surge in unusual bulk orders, with some noting that buyers showed little concern for price, sometimes purchasing overpriced titles. Sellers speculated that these patterns could indicate AI companies with large budgets acquiring books for training purposes. ISBNdb’s now-removed pages argued that buying used books at scale does not deprive creators of income, since these books have already fulfilled their financial obligations to authors.
The issue of printed books being used for AI training has drawn wider attention. In January, book authors filed a copyright lawsuit against Anthropic, alleging the company planned to obtain and scan millions of printed books, destroying them in the process. The Washington Post reported that Anthropic sourced books from Better World Books, a major reseller of used titles. ISBNdb’s own marketing materials acknowledged the reputational risk, stating that being seen destroying millions of books would be damaging for any AI company.
Interest in sourcing pre-2022 printed books for AI training has been rising, as highlighted in a recent report on the surge in demand for uncontaminated training data.
