AI’s Quiet Appetite: The Secondhand Book Boom Explained
In the last months, independent booksellers like Stuart Manley of Barter Books in Northumberland have reported “mega orders” that would normally take a week to fulfil in just one wave. The rush has raised a ripple of questions across the industry: who wants so much book stock, and why are some titles being destroyed after sale?
The answer may lie outside traditional markets. In 2025 a U.S. judge ruled that using books purchased in bulk to train AI models does not violate copyright, citing the transformation of copyrighted text in machine‑learning processes. The ruling came after a lawsuit from three authors against AI firm Anthropic. In its defense, the company said it destroyed and archived destroyed copies on a large scale, a practice dubbed “Project Panama.”
Project Panama, revealed in court documents, describes an ambition to “destructively scan all the books in the world.” The process involves shipping physical books to facilities where each spine is stripped, pages are digitised quickly to harvest data, and the remaining case is recycled. The variety of books—from obscure Latin treatises to popular Westerns—speaks to the breadth of data AI needs to improve language models.
The practice has sparked a debate among sellers. Book owners praise the influx of cash for long‑unsold titles, yet many worry that rare or one‑of‑a‑kind editions might be lost forever. Edinburgh shop owner Derek Walker says there is a fine line: destroying a novel with only one surviving copy is “a different kind of loss” from wiping out a copy of a bestseller that has already saturated the market.
Besides the legal debate, the environment and consumer rights angle surfaces. If entire collections are being shredded, the question arises whether publishers and authors should be compensated for the data value of their works. Meanwhile, libraries and archival institutions warn that unchecked destructive scanning could erase rare manuscripts that serve cultural heritage.
For readers and authors, the emerging trend underscores the shift from tangible to digital abundance. The next chapter in the relationship between books and AI will likely hinge on new legislation and transparent industry practices to balance profit, intellectual property, and cultural preservation.
