Secondhand book sales are booming.
Independent booksellers have logged massive orders that surpass a month’s worth. Stuart Manley of Barter Books in Northumberland sold a bulk that would otherwise take a week, sending him perplexed and curious.
In 2025, a U.S. court ruled that using such books for AI training was not a copyright breach. Judge William Alsup said the books’ use was “exceedingly transformative,” giving Anthropic’s Claude chatbot clearance. When the papers were released, notes about destructive scanning and a project dubbed *Project Panama* surfaced.
Destructive scanning means books are shipped to sites for rapid digitisation. Spines are removed, pages are scanned, and remaining bits are recycled. This process, intended to feed large language models, has left booksellers uneasy because some rare editions vanish with no trace.
Booksellers face a dilemma. Derek Walker of McNaughtan’s points out that while a 100‑copy academic text may not be a huge loss, destroying a sole 18th‑century edition is far more consequential. Yet many titles languish on shelves; Manley sees a chance to profit and recycle.
While AI may benefit from fresh, diverse text, the practice raises legal and ethical questions in the UK, where permission from copyright holders is required, according to Oxford professor Emily Hudson.
Want to explore AI and tech trends? Sign up for our Tech Decoded newsletter.



















