In brief
Large numbers of used books are being shipped under bulk orders to unknown destinations, a pattern that has caught the eye of independent booksellers worldwide. Legal analysis points to a U.S. court ruling that qualifies the use of purchased works for AI training as transformative, sparking speculation that firms like Anthropic may be quietly building enormous datasets. The practice—dubbed "Project Panama"—has led to concerns over destructive scanning and the possible loss of rare titles. While AI businesses claim the process is industry-standard and not destructive, booksellers and researchers warn of significant intellectual and cultural implications. The debate has highlighted differences in copyright law between the U.S. and the U.K., and has called for clearer regulations around the use of copyrighted material in AI training.
Rising Bulk Orders and a Hidden Market
Independent sellers around the world have reported a sudden spike in large orders of secondhand books. Stuart Manley of Barter Books in Northumberland noted a single purchase that matched seven days’ worth of sales in one week. This pattern, unexpected after 30 years in the trade, has led booksellers to wonder about the ultimate destination of the books.
Court Ruling that Could Explain the Trend
In 2025, a U.S. district judge ruled that using purchased books to train AI models was “exceedingly transformative” and therefore not a copyright infringement. The decision arose from a lawsuit brought against Anthropic by three authors. Judge William Alsup granted Anthropic broad permission to use the works, a ruling that opened the door for large‑scale data acquisition.
According to court documents, Anthropic’s project—codenamed “Project Panama”—was described as an intent to “destructively scan all the books in the world.” Destructive scanning involves shipping books to facilities where they are stripped of spines, photographed, and then the physical book is recycled or destroyed to accelerate digitisation at industrial scale.
Inside the AI Industry’s Data Practices
Anthropic’s spokesperson emphasised that the company buys books only for training purposes, not to destroy rare editions. The firm says its data acquisition processes are standard across the industry, which includes a mix of publicly available web data, commercially purchased datasets, and internally generated material.
Booksellers, however, are uneasy. Some titles—once thought lost or rare—are being shipped back for scanning. “I’ve carried on books for 20 years that never sold until now,” says Stuart Manley, reflecting on the tension between economic opportunity and cultural loss.
Legal and Ethical Cross‑Continent Perspectives
Professor Emily Hudson, an intellectual property specialist at Oxford, explained that U.K. copyright law requires permission from authors for copying and training. In contrast, the U.S. court described the use as transformative, a difference that has led to international debate over the scope of AI use.
For booksellers, the key question is not whether a bulk order is legal, but whether it represents a moral compromise. Derek Walker of Edinburgh’s McNaughtan’s says “if we sell the single surviving copy of an 18th‑century edition for destruction, it would be a grave loss for cultural heritage.”
It is also worth noting that some AI firms argue that the bulk of scanned material is from works already widely distributed, implying that the net loss to society may be negligible. Yet the sheer volume of scanning calls for stricter oversight and transparent practices from industry leaders.
Future Questions and Call for Regulation
The trend invites several questions: How should governments steer the balance between AI innovation and protection of literary heritage? What frameworks will ensure transparent reporting of data acquisition? And how can authors and booksellers secure a fair deal when the market is shifting beneath them?
Until clearer policies are enacted, the global community faces a dilemma. Is the rise of AI borrowing from the collective shelf a boon for technological advancement, or is it eroding the very cultural artifacts that inspire those innovations?
