AI Giants Accused of Destroying Physical Books to Feed Training Models

Major AI developers are engaging in the systematic destruction of physical books to secure high-quality training data, according to investigative reports. The practice involves purchasing millions of volumes, removing their spines for high-speed scanning, and shredding the originals to build proprietary datasets.

Anthropic reportedly operated a clandestine operation codenamed "Project Panama," which an internal memo described as an effort to "destructively scan all the books in the world." Amazon has also been linked to similar operations, with reports tracking books to Amazon processing facilities for destructive scanning and shredding. To facilitate these acquisitions, ISBNdb previously offered bulk book-acquisition services for LLM datasets, pitching the idea that the world's best training data is "sitting on a shelf."

The Drive for 'Unpolluted' Data

This shift toward physical media is driven by a plateau in available high-quality digital data. As AI-generated content—often called "AI slop"—floods the internet, developers are seeking "unpolluted" sources written before the proliferation of large language models, roughly prior to 2022. Destructive scanning is preferred over non-destructive methods because it is significantly faster and cheaper for processing millions of pages.

Legal incentives have further accelerated this trend. In the $1.5 billion copyright settlement Bartz v. Anthropic PBC, the court ruled that training AI on books constitutes fair use, but noted that storing pirated libraries does not. This ruling incentivized companies to purchase physical copies of books to avoid the legal risks associated with digital piracy while still benefiting from the content.

The Risk of Knowledge Monopolies

Critics argue that this practice creates a dangerous "knowledge monopoly." By destroying physical copies—including rare editions—companies ensure that the only remaining versions of certain works exist on private corporate servers. This effectively erases the physical public record and prevents independent researchers, competing developers, and the general public from accessing the original material.

In response, the open library Anna's Archive has issued a global call for volunteers to scan and upload physical books. The organization warns that knowledge is being "permanently monopolized on private servers" and urges a race to preserve human knowledge before it is lost to corporate interests.

Regulatory Fallout

The controversy has already reached regulators, with advocates filing complaints with the Federal Trade Commission (FTC). These complaints allege anticompetitive behavior and the intentional monopolization of rare texts to gain a market advantage in AI capabilities. What remains to be seen is whether the FTC will view the destruction of physical cultural heritage as a violation of antitrust laws or a permissible byproduct of corporate data acquisition.