AI firms purchase and shred books to train models

Artificial intelligence companies are acquiring large numbers of physical books, including rare and out-of-print editions, to scan their contents for AI training data and subsequently destroying the original copies, according to recent reports and court documents. The practice has ignited a debate over cultural preservation, copyright law, and the ethics of sacrificing physical artifacts for technological advancement.

The controversy centers on a process where companies purchase books in bulk, remove their bindings with industrial cutting machines, scan the pages, and then discard the physical volumes. Court filings from a copyright case involving AI company Anthropic detailed an internal initiative, known as Project Panama, in which the company spent millions of dollars to buy and destructively scan millions of printed books to expand the training dataset for its Claude chatbot. A federal judge ruled in June 2025 that this specific practice of scanning legally purchased books for internal AI training constituted fair use, as the digital copies replaced the owned physical ones without increasing the total number of copies in circulation.

The demand is driven by the search for...