02/08/2026
Millions of printed works are sent to service providers, cut open, and scanned page by page. In the end, only a digital file remains—and a new copyright dispute.
Books are considered cultural heritage. For AI companies, however, they are above all valuable training material. To develop their language models, firms like Anthropic purchase millions of physical books in bulk, often sourcing from European and international book sellers. The works are not resold or archived afterward. Instead, service providers slice off the book spines, digitize each page using high-speed scanners, and dispose of the paper stacks in recycling.
Court documents from a copyright lawsuit against US company Anthropic reveal how systematic this approach has become. The documents, made public in January 2026, provide rare insight into the AI industry's race for high-quality training data. Under the codename "Project Panama," Anthropic systematically purchased millions of books in 2024—including rare and out-of-print titles—and had them dismantled by hydraulic cutting machines. An internal planning document described the effort plainly: "Project Panama is our effort to destructively scan all the books in the world."
The company acknowledged that this method was deliberately kept hidden, aware of how negatively such practices would be perceived. In June 2025, a federal judge determined that scanning legally purchased books may constitute "fair use" under U.S. copyright law—a transformative use of the original works. However, Anthropic later paid a settlement of $1.5 billion for prior allegations involving the use of pirated books.
This practice raises important questions: What becomes of our cultural diversity when the last physical copies of rare or out-of-print titles are destroyed? While courts clarify the legal status, the ethical debate over the price culture and the book industry pay for artificial intelligence development continues to grow.
Images are generated by AI and for demonstration purposes only.
Source: Bartz v. Anthropic settlement (2025), Washington Post reporting on Project Panama (January 2026), court documents unsealed January 2026.
