The world's biggest artificial intelligence companies are reportedly buying millions of antiquarian books to be scanned for training large language models before destroying the physical copies, according to reports on the industry's data collection practices.

Among them is AI firm Anthropic, the developer of Claude, which reportedly spent tens of millions of dollars acquiring millions of books under an internal initiative codenamed "Project Panama."

The project involved cutting off book spines with hydraulic machines, scanning each page using industrial imaging equipment, and discarding the books after digitization.

Court documents unsealed earlier this year revealed that Anthropic recognized the project's controversial nature. An internal planning document described the initiative as an effort to "destructively scan all the books in the world" while noting that the company did not want it to become public.

Anthropic's scanning of legally purchased physical books was previously ruled to be fair use under U.S. law, with the court citing the first-sale doctrine, which allows buyers to dispose of purchased property as they choose. The ruling has contributed to a growing market for second-hand books acquired specifically for AI training.

The ruling came despite Anthropic's recent agreement to pay $1.5 billion to settle a class-action lawsuit over allegations that it used pirated books to train its AI models.