AI Companies Face FTC Investigation for Destroying Millions of Books After Scanning

AI Companies Accused of Systematic Book Destruction

A coalition of 18 civil society groups has formally requested the Federal Trade Commission to investigate AI companies for allegedly purchasing physical books in bulk, digitizing their contents for model training, and then permanently destroying the originals. The Demand Progress Education Fund, Consumer Federation of America, and Institute for Local Self-Reliance are among the organizations pushing for regulatory scrutiny of what they characterize as hoard-and-destroy practices that could constitute an unfair method of competition1

1

3

3

.

The controversy intensified after court documents in Bartz v. Anthropic PBC revealed details of "Project Panama," Anthropic's internal initiative described in a 2024 memo as "our effort to destructively scan all the books in the world." The company deliberately used a codename and instructed employees to avoid discussing the project publicly, stating the operation "should not be shared with anyone outside Anthropic"1

1

. Anthropic is not alone in this practice—a recent 404 Media investigation reported that Amazon has also been participating in book scanning and shredding operations4

4

.

Why Older Books Matter for AI Training Data

AI companies are specifically targeting pre-2022 works because these physical books are guaranteed not to contain AI-generated text, which has increasingly contaminated recent written material. Older books also tend to be better edited, making them more valuable for training large language models2

2

5

5

. After scanning, companies destroy the physical copies to avoid storage costs, but this practice raises serious questions about whether some rare books are being eliminated forever.

The advocacy groups' letter emphasizes a critical unknown: "What the public record does not establish—and cannot, from the outside—is how often the destroyed physical books are the last or among the last surviving copies of a given work"1

1

. In some cases, destroyed books may represent the final surviving copies of original works, permanently removing them from public access4

4

.

Anticompetitive Behavior and Barriers to Market Entry

The civil society groups are framing their FTC complaint primarily around competition concerns rather than copyright issues. Their letter argues that book destruction constitutes anticompetitive behavior designed to "create an insurmountable systemic moat around AI incumbents"3

3

. By permanently removing nonrenewable resources from broader access, AI companies are "starving the market" of critical source materials that competitors need.

Kate Oh, special advisor to the Demand Progress Education Fund, stated: "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"1

1

. The practice creates barriers to market entry by raising rival companies' costs and denying startups access to essential training materials3

3

5

5

.

Industry Backlash and Public Relations Concerns

The book destruction controversy has become a significant public relations headache for AI companies, partly because of the disturbing historical associations with book burning under authoritarian regimes, and partly due to broader backlash against AI firms for "pillaging public resources in pursuit of private gain"1

1

. The practice has spooked even AI devotees, drawing uncomfortable parallels to historical acts of cultural destruction2

2

.

Intermediaries are feeling the pressure. ISBNdb, which reportedly helped broker book acquisitions for destructive scanning, recently disavowed the practice and removed its landing page titled "Printed Books Sourcing for Your AI LLMs Dataset Needs," noting it was "exploring demand" but has "chosen to pivot away from that direction"1

1

.

Implications for Knowledge Access and AI Development

The advocacy groups warn that these practices 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"4

4

. This commodification of knowledge stands in stark contrast to OpenAI CEO Sam Altman's previous description of knowledge as a utility that could be sold "like electricity and water"—a proposition that appears far less wholesome when it involves wholesale destruction of books that remain accessible through libraries or at low cost2

2

5

5

.

The groups are asking the FTC to determine the scale of book destruction practices and investigate whether they violate Section 5 of the FTC Act, which prohibits unfair or deceptive acts in commerce4

4

. They emphasize they don't want to restrict AI model training itself, but rather focus on stopping the destruction of existing works and potentially intervening before AI companies build market dominance on this practice3

3

. The FTC investigation request represents a shift in the AI training data debate from copyright law into competition policy, potentially opening new regulatory scrutiny of how dominant AI firms eliminate resources their rivals need to compete.