The artificial intelligence industry, having already stripped the open web of its bark and devoured every morsel of its illegally acquired libraries, has developed a new appetite. Millions of physical books are now being bought in bulk and fed into a process that the industry nervously calls “destructive scanning”.
Data acquisition increasingly involves cutting the spines from millions of books, scanning them and then shredding and discarding them afterward.
The service comes courtesy of a company called ISBNdb. They promise to obtain “up to one million titles per order.” They specifically tout “older, rare and specialist volumes” as premium training data because books from the pre-2022 era are “structurally guaranteed” to be free of AI contamination. As AI starts to feed on itself into a delirious loop, information from the old world becomes more valuable to the momentum and continuation of large language models.
Read more about AI and the arts from Derek McArthur:
- A prestigious literary competition is being accused of rewarding AI-written works
- Replace all artists? The strange logic of the AI art defenders
- A rapper is pretending his new viral hit isn't a blatant AI creation
- OpenAI’s Sora was going to destroy Hollywood... until it came abruptly crashing down
- Can we pass a law banning the sale of Highland cow AI art?
- A nonsensical AI-generated mural is the last thing Glasgow ever needs
- This film veteran was stunned when he asked AI to give him ideas
The company's marketing material is blunt about the reality of what happens to the books. "'AI company destroys two million books' is not a headline that generates sympathy," its website states. It offers non-disclosure agreements to buyers and advises them to frame the work as "digital preservation".
We are talking about books that have survived through the many scourges of history and have become rare due to their life as an obscurity before the information age.
Small booksellers across Europe and the United States have noticed a sudden surge in bulk orders for rare and out-of-print titles. One bookseller told 404 Media that he went from selling about 20 books a week to hundreds, including hard-to-find foreign language works. "It benefits me financially as well as by clearing out old inventory that is otherwise unlikely to sell," he said. "On the other hand, I don't like the end-use, and I don't like that uncommon books are being pulped."
Okay, and on the other hand, these things cannot be replaced. Once a final non-digitised copy of a book is shredded, it’s gone forever. And the version that is digitised becomes at the mercy of AI interpretation rather than a straight, accurate representation of the work.
Anthropic, the company behind the chatbot Claude, has been a major driver of this practice. Internal documents made public during a copyright lawsuit revealed a program called “Project Panama” launched in 2024. Court filings described an industrial-scale operation using a "hydraulic powered cutting machine" to slice the spines off millions of books, which were then fed through "high speed, high quality, production level scanners". An internal memo called it "our effort to disruptively scan all the world's books" and noted, "we do not want it to be known that we are pursuing this project." Anthropic hired the former head of Google Books specifically to help acquire what the company described as "all the books in the world".
The legal framework enabling this comes from a federal judge's ruling in the Anthropic copyright case. The court determined that by making a digital copy and destroying the physical original in a one-for-one transfer, the process qualifies as "transformative" and falls under fair use protections. The reasoning is that since only one copy exists at a time, it's not traditional piracy. This ruling sets a precedent that allows the practice to wildly accelerate.
Read more from Derek McArthur:
- Does the Mack need to be faithfully rebuilt or are we just scared of our own future?
- Movie studios are circling like sharks to buy the biggest online review platform
- This year’s Cannes came and went. Is cinema in danger of becoming a niche hobby?
- Swapping the boss won't put back together a broken Creative Scotland
Rare booksellers across multiple countries are now receiving similar suspicious purchase requests. In the Netherlands, an antiquarian bookseller received an email from a Singapore-based company claiming to be working on "a new project focused on collecting books in multiple languages". The attached list contained 3,000 English-language titles organized by ISBN number, ranging from folklore and fairytales to technical manuals and science texts. Similar reports have emerged from Switzerland, Spain, and Germany. The booksellers involved concluded the books were destined for AI training because no collector or reseller would need such a random assortment of obscure titles.
So who will be the knight in shining armour and stop such a senseless and wasteful practice? Elon Musk stated on his social media platform X that he instructed his own AI team "to preserve any rare books in a library and scan them the hard way vs just cutting off the spine and scanning". Thanks for that, Elon, but it’s hard to take the moral right of someone seriously when their own AI model generated child sexual abuse material.
Destroying books is very efficient for AI companies and, according to the courts, legal. But the cost is a permanent loss that carries historical and cultural significance beyond words printed on the page. The books that survive this process will live the rest of days as mindless data points in a model, their physical forms long turned to pulp. Unless the legal framework changes or public pressure forces companies to preserve the originals, the Library of Alexandria will burn yet again.
Derek McArthur is an artist and arts writer specialising in cinema and culture. He writes a weekly arts column for The Herald.