Booksellers around the world are raising concerns that AI companies are buying rare books in bulk, scanning their pages, and then destroying the physical copies. The practice appears to be part of a race among tech firms to train their AI models on high-quality, long-form text that only exists in books.
The suspicion first gained attention when a Dutch bookseller received a request for 3,000 copies of a rare title. According to a report shared on Reddit, the bookseller initially thought the request was "spam or phishing." Instead, it appears AI companies are scanning and destroying books to build training datasets.
How AI Companies Destroy Books to Train Models
The process is straightforward but troubling for book lovers. AI firms buy books in large quantities, remove their bindings, scan the pages, and then discard the originals. As Morning Brew reported, this method is the cheapest and fastest way to digitize books at scale.
For AI companies, the goal is clear. They need engaging, high-quality long-form texts to advance their models, and books are a rich source of that material. But for booksellers and collectors, the destruction of physical copies is hard to accept.
Booksellers Call Destruction "Horrific"
The concern is not limited to one region. Australian booksellers have publicly raised the alarm about the practice. According to The Guardian, they describe the destruction of rare titles to feed AI as "horrific."
"More than just objects" — Australian booksellers on rare titles
Booksellers argue that rare books carry cultural and historical value that cannot be replaced. Once a physical copy is destroyed, it is gone forever, even if its contents have been digitized.
Reports of Shredding and Destruction
The scale of the practice may be larger than currently reported. As Ars Technica noted, a Telegraph report accused Silicon Valley of destroying millions of rare books and "shredding the originals." A separate report from 404 Media also covered the issue.
Book lovers fear the practice is happening on a grander scale than what has been documented so far. The combination of high demand for training data and the relative ease of destroying books makes it likely that more rare editions will be lost.
Our Take: The Cost of AI Progress
To put it plainly, this is a troubling trade-off. AI companies are building powerful tools that could benefit society in many ways. But destroying rare books to do it feels like a step too far.
These books are not just data. They are pieces of history, artifacts of how people once understood the world. Scanning them and throwing away the originals saves time and money, but it comes at a cultural cost that is hard to measure.
We think AI companies should find better ways to digitize books without destroying them. If they cannot, they should at least be transparent about how many books they are buying and what they are doing with the physical copies. Booksellers and collectors deserve to know what is happening to the rare titles they care about.
The suspicion among booksellers is not just about lost inventory. It is about losing something irreplaceable. That is a price that should not be paid quietly.
