A coalition of American civil society groups on Friday called for a federal investigation into reports that artificial intelligence giants are mass-purchasing physical books, scanning them to train their AI models, and destroying them without regard for historical artifacts or the public good.
In a letter to the Federal Trade Commission (FTC), the coalition notes that "at least one top AI company, Anthropic, is known to have acquired millions of print books in bulk, scanned them into digital form, and discarded the physical originals." The letter notes that "this process leaves only the digitized text in the company's proprietary database for use in training its large language models behind products such as Claude."
The coalition, led by the Demand Progress Education Fund, pointed to an internal Anthropic memo that was made public as part of a copyright lawsuit against the company. The memo is headlined "Project Panama," which is described as "our effort to destructively scan all the books in the world."
"We use a 'soft codename' for it because we don't want it to be known that we are working on this," the memo states. "This document is visible to all Anthropic employees, but you should avoid talking about it in public areas, and the fact that we are working on this should not be shared with anyone outside Anthropic."
At one point, Anthropic reportedly considered approaching US libraries such as the New York Public Library for its massive book-consumption campaign. One court document describes how an Anthropic vendor's "hydraulic powered cutting machine" would "neatly cut" books for scanning. The scanning company would then "schedule with the recycling company to pick up the completed books," according to the filing.
Amazon is also reportedly engaged in a large book-buying operation, under which employees "cut the bindings off in order to scan the books more quickly."
"The AI industry is acquiring books on vast commercial scales to scan and destroy, all to keep that scanned text in a private database walled off from the world and subject to corporate censorship and manipulation."
The new letter to the FTC observes that "proxies and intermediaries now function as sourcing partners for AI companies, offering the buyers anonymity and brokering bulk book orders—including rare, out-of-print titles with few surviving copies." 404 Media reported last month that ISBNdb, which markets itself as "the world's largest book database," is offering "high-volume book acquisition services for AI companies."
"Books published before the rise of generative AI (such as Claude, ChatGPT, and Gemini) command a premium because they are free of the AI-sourced text common in today's writing," the new letter adds.
The coalition argues that AI giants' mass book buying and subsequent destruction—dubbed the "hoard and destroy" practice—is part of an effort to "build and entrench their market dominance" by "corner[ing] the essential inputs of AI development."
"We ask the FTC to use its investigative authorities, including potentially under Section 6(b) of the FTC Act, to determine whether and to what extent this practice forecloses competing AI developers and the public from a scarce, non-reproducible input and whether it constitutes an unfair method of competition under Section 5 of the FTC Act," the groups write.
"Through their practice of permanently destroying books en masse and thus removing those non-renewable resources from broader access, AI companies are engineering a future where only the wealthiest incumbents can build high-quality AI models and operate as the sole holders of humanity's written works—after having destroyed the originals to get there."
Kate Oh, special adviser to the Demand Progress Education Fund, said in a statement Friday that "the secretive and reckless way that major AI companies like Anthropic and Amazon are acting shows that there is real smoke here that the FTC needs to investigate."
"There is credible reporting that the AI industry is acquiring books on vast commercial scales to scan and destroy, all to keep that scanned text in a private database walled off from the world and subject to corporate censorship and manipulation," Oh added.
