U.S. civil society groups have called for an investigation into the practice of major tech companies buying up and destroying books on a massive scale for artificial intelligence (AI) training.

Eighteen organizations, including the U.S. civic group Demand Progress Education Fund and the Consumer Federation of America (CFA), announced on the 21st (local time) that they sent a joint letter to Federal Trade Commission (FTC) Chair Andrew Ferguson and Commissioner Melissa Holyoak urging an immediate investigation into such practices.

In the letter, they argued that AI companies purchasing paper books in bulk, scanning them, and then permanently discarding the original physical copies constitutes unfair market dominance.

In particular, they argued that this type of book destruction amounts to input foreclosure that artificially blocks the entry of latecomer startups, demanding that the FTC launch an investigation by applying federal laws prohibiting unfair competition.

"By permanently destroying books in large quantities, AI companies are blocking broader public access to these resources," they criticized, adding, "This has allowed only the wealthiest incumbent companies to build high-quality AI models."

They also criticized that "they are creating a future where they destroy original copyrighted works to remain the sole exclusive record-keepers of humanity."

The books targeted for AI training are mostly works written before AI became widespread, making them finite and scarce human resources, they pointed out.

Anthropic built its AI training data through a so-called Project Panama, in which it purchased millions of books, sliced off their bound spines, scanned them, and recycled the destroyed books as waste paper.

Amazon is also known to have used a similar method to conduct AI model training.

The British daily The Guardian recently reported that there had been massive buying sprees of used books—seemingly selected by machines—at secondhand bookstores in North America, Europe, and Australia.

The information technology specialized media outlet 404 Media reported that after attaching tracking devices to books at a bookstore that received a bulk order of 1,000 books, it was confirmed that the books headed to an Amazon fulfillment center in Las Vegas.

Previously, a U.S. federal court ruled in a copyright lawsuit filed by authors against Anthropic that legally purchasing books and using them for AI training falls under fair use and does not constitute a copyright infringement.

However, civic groups maintain that even if it does not violate copyright law, it amounts to unfair competition.

Meanwhile, Anthropic settled its lawsuit with authors last month under the condition of paying 1.5 billion dollars (approx. 2 trillion KRW), the largest settlement in the history of U.S. copyright lawsuits.

(Photo: Capture of evidentiary documents submitted to the U.S. District Court for the Northern District of California (Case No. 3:24-cv-05417), Yonhap News)