HR EN DE
NEWS SPORT BIZNIS SCENA LIFESTYLE TECH

Project Panama: How Anthropic Secretly Destroyed Millions of Books

Unsealed court documents reveal shocking details of a covert operation where books were sliced and scanned, then turned into toilet paper. The case ended in the largest author settlement in history.

Foto: Pavel Danilyuk na Pexels
Summary
  • Anthropic destroyed millions of books in the secret Project Panama to scan their content for training the AI model Claude.
  • The company agreed to pay $1.5 billion in the largest author settlement in history, covering approximately 500,000 works.
  • Co-founder Ben Mann had been downloading pirated books from the shadow library LibGen since 2021 and shared them with colleagues.
  • Meta, OpenAI, Google, and Microsoft also face similar lawsuits from authors over the use of copyrighted works for AI training.

In a scene reminiscent of a Ray Bradbury dystopian novel, American AI company Anthropic conducted a secret project codenamed Panama, during which machines in warehouses across the United States sliced book spines like salami, scanned their contents, and then sent the destroyed paper to recycling. According to a Euronews report citing unsealed court documents, the goal was to "destructively scan all the books in the world" to train the AI model Claude.

An internal Anthropic document, made public in late July 2026, explicitly states: "We don't want anyone to know we're working on this." The project was launched in early 2024, with executives seeking books older than those available online, believing they would teach Claude "how to write well," rather than mimic "low-quality internet speech."

How the Book Destruction Factory Worked

Anthropic purchased thousands of books at once from multiple suppliers, including Better World Books and the UK-based World of Books. According to N1 Hrvatska, the company acquired hundreds of thousands of books. Euronews, meanwhile, reports that an external supplier, which eventually worked with Anthropic, was tasked with "finding an experienced document scanning service provider to convert 500,000 to 2 million books over a six-month period."

After purchase, the books were handed to a subcontractor who used a hydraulic machine to cut off spines and slice pages one by one. The pages were then scanned with high-speed industrial devices, and the digital content was uploaded to the AI training system in milliseconds. The destroyed paper ended up in recycling plants where it was turned into products like toilet paper and cardboard boxes.

Pirate Libraries and the LibGen Shadow Library

Court filings reveal that destroying physical books was not the only way Anthropic obtained copyrighted content. Co-founder Ben Mann had already downloaded a collection of copyrighted material from the shadow library LibGen in June 2021, and the filing includes a screenshot of Mann's browser during the download of pirated books.

In 2021, Mann shared a link to the Pirate Library Mirror, a complete catalog of illegally obtained books, with other Anthropic employees, commenting: "Just in time!!!"

Authors' Lawsuit and Historic Settlement

Project Panama only came to light through thousands of pages of court documents unsealed in July 2026. Back in 2024, a group of authors, led by writer Andrea Bartz and publicists Charles Graeber and Kirk Wallace Johnson, filed a class-action lawsuit against Anthropic, claiming the company violated copyright laws by using pirated books to train its AI model.

Last month, Anthropic, valued at an estimated $965 billion (about €875 billion), agreed to pay $1.5 billion (about €1.3 billion) to settle, covering approximately 500,000 eligible works. "This is the largest known author settlement in history," said Justin Nelson, the authors' attorney.

A June 2025 ruling found that using legally purchased books was protected under the fair use doctrine in U.S. law, but drew a distinction between purchased books and the millions of previously pirated ones. "The June 2025 ruling remains intact. The issue we resolved was about how some materials were obtained, not whether we can use them for development [of AI]," Aparna Sridhar, deputy general counsel at Anthropic, told the Washington Post.

AI Industry Under Scrutiny

The Anthropic case is the first in a series of lawsuits against AI companies to end in a settlement. Meta, OpenAI, Google, and Microsoft also face similar copyright infringement lawsuits from authors. Several authors who declined to participate in the class settlement are still pursuing individual cases against Anthropic.

The documents also revealed the mindset at the top of the AI industry, suggesting that Anthropic executives may have believed they would get away with it, and even if they didn't, that the end justified the means. For many observers, the unsealed documents represent a more lasting judgment, shattering the illusion that the world's largest tech companies understand cultural value or respect the sanctity of human creativity.

FAQ
What was Project Panama? +
A secret project by AI company Anthropic in which hundreds of thousands of purchased books were destroyed and scanned to use their content for training the AI model Claude.
How much did Anthropic pay for the settlement with authors? +
Anthropic agreed to pay $1.5 billion (about €1.3 billion), making it the largest known author settlement in history.
Are other AI companies facing similar lawsuits? +
Yes, Meta, OpenAI, Google, and Microsoft also face lawsuits from authors claiming their works were illegally used to train AI models.
What did the court rule about using books for AI training? +
In June 2025, the court ruled that using legally purchased books is fair use, but drew a distinction between purchased books and the millions of pirated books Anthropic had previously used.

Log in

You need to log in or register to comment.

Comments (0)
No comments yet. Be the first!
Traži
Popularno
Nedavno pretraživano
helsinški sporazum
liga prvaka
digitalni mediji
Login
Home
Prati nas na Googleu
Categories