Something mysterious has been happening in the world of secondhand books. For the last few months, independent booksellers across the globe have reported a strange pattern: massive bulk orders that dwarf everyday sales.
Stuart Manley of Barter Books in Northumberland, for example, sold only two to three thousand books in a normal week. A single Canadian company, however, bought an order that equalled what he would normally move in seven days. He says he has "never seen the like of this after 30 years in the second‑hand book trade.”
Other sellers echo this experience, noting an influx of books ranging from obscure Latin texts to cowboy novels, with no clear final destination. Many suspect the ultimate buyer is not avid readers but artificial‑intelligence training programs.
In 2025, a U.S. judge ruled that using books purchased in this manner to train AI software does not violate U.S. copyright law. Judge William Alsup noted that Anthropic’s use was "exceedingly transformative", allowing it under the law. Court filings now expose the company’s internal project, known as "Project Panama", aimed at "destructively scanning all the books in the world." The process involves shipping books to facilities where they are rapidly digitised—often with the spine cut—before the remains are recycled.
Many booksellers feel uneasy about the idea of books being destroyed for AI training. Derek Walker of McNaughtan’s in Edinburgh reflects on the weight of a single 18th‑century edition that could be lost if bought for pulping. Yet others are willing to accept the trade, stating "the world no longer needs five million copies of The Da Vinci Code."
Alexis L. from a London shop notes the surge offers a double‑edged sword: while it brings in extra income, it also raises ethical questions about stealing rare copies for the sake of a machine’s data. The debate continues as stakeholders weigh economic benefits against the cultural loss of physical books.
For more insight on how AI works and its implications, visit Tech Decoded newsletter.
