Project Panama Came to Light Only After Authors Took the AI Company to Court
Millions of physical books passed through industrial cutting machines so Anthropic could build a digital library for its Claude artificial intelligence models. The bindings came off. The pages were trimmed, scanned and converted into searchable files. The paper originals were then discarded.
Most people never heard about this operation.
How Project Panama Became Public
Anthropic called the effort Project Panama. According to internal records reviewed by The Washington Post, the company began buying books in bulk in early 2024, often by the tens of thousands. A contractor proposed converting between 500,000 and 2 million books within six months. Anthropic ultimately acquired millions.
The company did not announce the program. One internal planning document described it as an effort to “destructively scan all the books in the world” and said Anthropic did not want the work publicly known.
The details surfaced because authors sued Anthropic over its use of copyrighted books. The litigation produced thousands of pages of internal records, some of which were later unsealed. Journalists examining those records reconstructed the purchasing and scanning operation.
(Below) CBS News interviews 404 Media co-founder Emanuel Maiberg about evidence that Anthropic is not alone in destroying books for AI. An investigation tracked rare books to an Amazon facility, where books are scanned and destroyed.
Why Destroying Books Matters in Court
The lawsuit exposed two different book-acquisition methods that are easy to confuse. Anthropic had downloaded millions of books from unauthorized online libraries. Separately, it bought physical books, destroyed them during scanning and retained digital replacements.
A federal judge treated those methods differently. In a June 2025 court ruling, Judge William Alsup found that converting a legally purchased print book into one private digital replacement qualified as fair use under the circumstances. Anthropic’s pirated library did not receive the same protection. A $1.5 billion settlement approved in July 2026 covered pirated works, not the books Anthropic purchased.
Was Anthropic Legally Required to Destroy the Purchased Books?
No law ordered the company to cut and shred them. Destruction was partly a practical choice. Removing bindings allowed pages to move quickly through industrial scanners.
Destruction also strengthened Anthropic’s legal position. The company characterized the process as a one-for-one format conversion: one physical copy disappeared as one private digital replacement was created. Selling or donating an intact book after scanning would have left two usable copies, potentially weakening that fair-use defense. The judge emphasized that Anthropic neither increased the number of copies nor distributed its digital replacements outside the company itself.
The ruling leaves Congress with a policy question: If destroying the original makes a company’s copyright argument stronger, has the law created an incentive to destroy books?
On August 21, a coalition of 18 public-interest organizations asked the Federal Trade Commission to investigate whether AI companies’ “hoard-and-destroy” practices suppress competition and permanently remove scarce books from public access.
Were the Books Rare or Out-of-Print?
Internal Anthropic documents presented in the court case show Project Panama prioritized “less common books.” Anthropic says that meant harder-to-find professional and reference works, not collectible or antiquarian editions. Without a public inventory, nobody outside the company can verify whether scarce copies went through the cutters.
An out-of-print book is not necessarily rare. Publishers often stop producing a title when plenty of used copies remain available.
Is Anthropic Still Destroying Books?
Public records describe the program’s operation beginning in 2024, but Anthropic has not provided a clear accounting as to whether the practice was shut down. The practice may still be going on.
The public still lacks basic answers. A private company quietly turned millions of physical books into proprietary training material. Court records revealed how Anthropic did it. They do not show whether the program continues. Nor do they identify which books disappeared from circulation. Why do AI companies, like Anthropic, want to keep the public from knowing?
0 Comments