Reports that rare books were being bought in bulk, scanned and destroyed for AI training have triggered a fierce backlash. Schools, libraries and booksellers were already operating with limited resources when unusually large orders began appearing, prompting concerns about who was behind them.
ISBNdb became a particular focus after appearing to offer a way for AI companies to obtain printed books from depositories. The book database has since removed the relevant material and is responding to the controversy in an update.
“We’ve seen the recent coverage about a marketing landing page on our site, and we understand the concern it raised,” ISBNdb wrote. “We don’t train AI models, and we never have. The page was a test of market interest; no such service was ever brought to life. We’ve taken the page down.”
What ISBNdb appeared to propose
A deleted landing page presented books as especially valuable AI training material, describing them as curated, peer-reviewed and domain-specific human knowledge that web crawling could not reproduce. It also suggested that printed books could be supplied for training purposes.
The page’s message was particularly damaging because it appeared while booksellers were receiving sudden bulk requests. The orders were suspected of coming from AI companies seeking cleaner material, while ISBNdb seemed positioned to connect those companies with book collections.
Why rare books are being scrutinised
The broader practice became public earlier this year after court documents revealed Anthropic’s “Project Panama”. The programme involved obtaining books, scanning them and then destroying the physical copies. Anthropic later settled with authors for $1.5 billion.
Unsealed documents indicated that the practice was regarded as legally defensible, despite its damaging public image, and that the company was prepared to spend heavily to keep it private. With that information now public, unusual orders for scarce books are receiving much closer attention.
One anonymous seller described the situation as financially useful because it cleared old stock that was unlikely to sell, but also expressed discomfort with the intended use and the pulping of uncommon titles.
Coverage of the book-scanning practice emerged in January, while the latest ISBNdb controversy has brought the issue back into public view. For readers interested in how technology changes the way information is preserved and accessed, RETROCADE’s arcade news and editorial coverage explores similar questions across gaming culture.
The search for cleaner training data
AI companies are increasingly looking for high-quality source material as low-grade web content, including text generated by other AI systems, threatens to create feedback loops and poorer results. Earlier this summer, A24 announced a partnership with Google to help train DeepMind, with the stated aim of improving media-generation systems.
The destruction of books may not be intended to conceal the activity or protect some uniquely valuable knowledge. The available Project Panama documents did not explain why the books were discarded. Cost is one likely explanation: scanning can be performed more quickly and cheaply when books are dismantled, with their spines removed to make the pages easier to capture before the remains are thrown away.
The dispute also highlights a familiar preservation principle for arcade enthusiasts: original material matters. Whether maintaining a classic cabinet or choosing an arcade parts and repair solution, the value often lies in keeping the underlying source intact rather than treating it as disposable.
For anyone planning a home setup, RETROCADE’s arcade machine buying guide offers a practical starting point, while DIY arcade kits provide another way to build around preservation and repair instead of replacement.
Leave a Reply