Booksellers suspect AI firms are buying and then destroying rare books
AI firms quietly bulk buying rare books face resistance from booksellers.
The rare book trade has always been a quiet business, conducted in the hush of archives and the polite competition of auction houses. Now it has become a front line in the AI data race, with booksellers reporting that anonymous buyers are acquiring scarce volumes in bulk, only to have them destroyed after scanning. The motive is not malice but efficiency: cutting the spine and feeding pages through a scanner is faster and cheaper than any conservation-minded alternative.
What makes this practice particularly striking is the asymmetry of value. A rare book holds cultural and monetary worth that appreciates over decades, while the AI company extracts a transient training signal that will be obsolete within a few model generations. The bookseller sees a unique artifact; the buyer sees a disposable input. Neither side is irrational, but their valuations are so divergent that the transaction feels like a kind of theft, even when the price is fair.
The destruction is not inevitable. Libraries and digitization projects have long shown that books can be scanned without being sacrificed, using overhead cameras and careful handling. The fact that some AI firms choose the faster, rougher method reveals a simple calculation: time is more valuable than preservation. In a race to train the next model, a few thousand rare books are a rounding error.
For the labor market, the story is a reminder that AI's appetite for data is not abstract. It has physical consequences, and the people who feel them first are often those in niche, analog professions—booksellers, archivists, and conservators. Their expertise is being bypassed, not because it is worthless, but because it is slow. The market is signaling that speed trumps craft, and the books are paying the price.
There is a quiet irony in the fact that the very texts AI companies seek—long-form, high-quality, human-made—are being consumed in a way that ensures they can never be read again. The models may learn to imitate the prose, but the original becomes ash. That is the trade-off, and it is being made without public debate, one rare book at a time.