Anthropic spent tens of millions of dollars buying books in bulk, slicing off their spines, and scrapping the pages after scanning
Anthropic, the American firm behind the Claude chatbot, bought millions of printed books, cut them apart, scanned the pages, and threw the remains away. The operation was revealed in court documents unsealed in January and reported by The Washington Post.
The company called it Project Panama. Internal planning papers described it as ‘our effort to destructively scan all the books in the world’. Staff were told the project ‘should not be shared with anyone outside Anthropic’, according to documents cited by the Dallas Express.
How the books were destroyed
Anthropic hired Tom Turvey, a former Google Books executive, in early 2024 to run the operation, Ars Technica reported. The firm bought books by the pallet from used-book wholesalers and spent tens of millions of dollars on purchases and scanning.
Hydraulic cutting machines sliced the spines off so that loose pages could pass through high-speed scanners. Destroying a book this way is faster and cheaper than scanning it intact, one page at a time. Once scanned, the pages were recycled or discarded.
Vendor proposals discussed processing 500,000 to 2 million books over roughly six months, according to Storyboard18. Books published before 2022 were prized because they predate the flood of AI-generated text online. That makes them cleaner sources of human writing.
What the court decided
The details came out through Bartz v Anthropic, a copyright case brought by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson.
In June 2025, US District Judge William Alsup ruled that training AI on legally bought books was fair use under American copyright law. His written ruling also held that Anthropic could lawfully turn purchased print books into digital copies, provided the physical books were destroyed and the files kept in-house.
Project Panama is our effort to destructively scan all the books in the world. — Anthropic
The judge ruled against Anthropic on a separate issue. The firm had downloaded millions of pirated books from shadow libraries such as LibGen. Those claims were settled for $1.5 billion, the largest copyright settlement in American history, covering around 500,000 works at roughly $3,000 each, according to the Authors Guild. The settlement was agreed in August 2025 and given preliminary approval in September 2025, legal analysts at Wolters Kluwer note. Anthropic also agreed to destroy the pirated files.
The hidden market for old books
A commercial market has grown up around feeding printed books to AI firms. ISBNdb, a commercial bibliographic data company, has expanded into helping clients buy books in bulk. Orders range from about 1,000 books to as many as one million at a time, with buyers’ identities protected by confidentiality agreements.
Google Books faced its own copyright battle more than a decade ago and survived it. The difference is that Google scanned books without destroying them. Anthropic’s approach leaves nothing behind but the digital file. Anthropic’s operation is now public. The identities of other AI firms buying books in bulk, and the total number of books destroyed across the industry, remain unknown.
