[REVISION HISTORY]

AI book digitization and destructive scanning controversy

Updated 11 times since CLSTR started tracking revisions of this situation.

What changed

2026-08-28 10:16 UTC → 2026-09-05 12:17 UTC · added ~~removed~~

In late July 2026, SpaceXAI announced it would use non-destructive scanning for rare books, contrasting with the high-speed, spine-cutting methods used by other firms. This follows reports into Anthropic’s ‘Project Panama,’ which involved purchasing large batches of books, cutting their spines for industrial-speed scanning, and destroying the paper copies. While a U.S. federal judge previously ruled that scanning legally purchased books can qualify as fair use, Anthropic later settled related lawsuits for $1.5 billion. By mid-August 2026, scrutiny intensified as booksellers in Europe, Australia, and North America reported unusual, large-scale orders of eclectic titles. Experts suggest these acquisitions aim to secure high-quality, human-written text to prevent ‘model collapse’—a phenomenon where AI models degrade by training on their own synthetic output. New investigations ~~have further detailed the scale of these operations. An investigation~~ by 404 Media used an Apple AirTag to track a shipment of books to an Amazon-owned warehouse, where employees reported scanning books as a primary task. Reports also indicate that facilities, such as an Amazon scanning site in Las Vegas, have been observed cutting spines to facilitate rapid digitization. ~~While the surge provides a financial windfall for some dealers, many express concern over~~ In early September 2026, the ~~permanent loss~~ pattern of ~~rare~~ unusual acquisitions expanded to include specialized academic content. Bookstores in Canada reported surges in bulk orders for older, technical volumes covering subjects such as medicine, mathematics, and physics. These orders often consist of random lists of titles that have historically ~~significant physical volumes.~~ remained unsold. In several documented cases, books ordered from Toronto-based retailers were shipped to a logistics company in Illinois, further fueling suspicions that AI companies are targeting specialized texts for training data.

Versions

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.