The AI Data Arms Race Takes a Dark Turn: Book Destruction Tactics Under Scrutiny

The U.S. Federal Trade Commission is facing mounting pressure to examine a controversial data-gathering strategy within the artificial intelligence industry. A coalition of advocacy groups has formally requested an investigation into whether major AI companies are engaging in anti-competitive practices by purchasing physical books, scanning them for training data, and then destroying the original copies.

Beyond Copyright: A New Form of Market Control?

According to court documents reported by The Washington Post, AI firm Anthropic spent millions to acquire books, remove their bindings for efficient scanning, and used the content to train its Claude model. Similar allegations have surfaced in copyright lawsuits against Google, Microsoft, and OpenAI.

The coalition's letter to the FTC, however, shifts the focus from intellectual property to market fairness. The central question posed is whether this "buy-and-destroy" methodology constitutes an unfair method of competition under U.S. law.

The Hidden Costs: Depleting Heritage and Stifling Innovation

The groups outline a twofold concern that extends far beyond typical business rivalry:

  • Irreversible Cultural Depletion: When AI companies become the final owners of physical books—especially rare or out-of-print editions—they risk permanently erasing a tangible piece of cultural heritage. While digital copies remain, the destruction of the original artifact represents a loss of historical context and preservation.
  • The Architecture of Data Advantage: A more systemic risk involves the structure of the AI market itself. By controlling and consuming finite data sources, industry leaders could:
    • Dramatically increase costs for competitors seeking similar-quality data
    • Deprive startups of the raw materials needed to train competitive models
    • Transform data access into an insurmountable barrier to entry

A Call for Targeted Intervention

The letter is careful to note that its signatories are not asking the FTC to restrict AI model training broadly. Instead, they urge regulators to specifically scrutinize the act of destroying existing works and to intervene before large tech firms can cement lasting market advantages through this practice.

This approach is framed not merely as aggressive data acquisition, but as a potential tool for building "systemic moats"—structural advantages that could lock in dominance for early movers while freezing out future competition.

The request places the FTC in a challenging position. The commission has historically balanced a business-friendly regulatory stance with vigilance against monopolistic practices in the tech sector. This new petition reframes the debate around AI governance, suggesting that when the pursuit of data leads to the physical eradication of cultural resources, regulators may need to draw new lines in the sand.