It sounds like a Ray Bradbury story come to life.
Somewhere in America’s sprawl of warehouses, book spines were sliced by machines like cuts of salami. Scanners then processed their contents in bulk. Within milliseconds, work that may have taken humans years to produce was uploaded into the digital ether, and later used to train AI models to replicate the voice and structure of our languages.
From there, the paper became waste, which was shipped off to recycling plants to be pulped and repurposed into products like toilet paper and cardboard boxes.
In a name that evokes both a covert operation and documents exposing financial sleight of hand, Project Panama was Anthropic’s secret plan “to destructively scan all the books in the world,” an internal planning document unsealed in legal filings revealed in late July.
“We don’t want it to be known that we are working on this,” the document said.
The revelation of Anthropic’s shadowy efforts – first reported by the Washington Post – has cast another pall on the AI arms race.
It has also offered one of the clearest looks yet at how one of the world’s biggest AI firms views the work of authors, artists, photographers and other creators.
How Project Panama fed the paper mill
Project Panama began in early 2024, as Anthropic executives sought books predating those available online that could teach its AI chatbot Claude “how to write well” instead of mimicking “low quality internet speak”, according to newly released documents.
The startup purchased tens of thousands of books at a time from multiple vendors, including Better World Books and UK-based World of Books.
It then outsourced the heavy work of scanning and destroying books to a vendor, which used a hydraulic machine to cut off the spines and slice pages one by one so they could be quickly copied and uploaded using industrial imaging equipment.
According to a vendor that ultimately worked with Anthropic, the company was “seeking an experienced document scanning services vendor to convert from 500,000 to two million books over a six-month period.”
Why books instead of the internet?
Internal planning documents suggest Anthropic executives believed the initiative would lead to higher quality communication and address AI model collapse – a kind of degradation that occurs when AI trains on text contaminated by its own output.
Unlike much of the modern web, books offer vast quantities of human-written and professionally edited literature yet to be spoiled by AI-generated content.
The challenge for the AI companies was a matter of how to obtain them. But the court filings suggest that Anthropic executives may have believed they could get away with it, and even if they couldn’t, that the end would justify the means.
According to one filing, Anthropic co-founder Ben Mann acquired a trove of copyright-protected content from a “shadow library” called LibGen in June 2021. The filing even included a screenshot of Mann’s web browser in the process of downloading pirated books.
The following year, Mann shared a link to the Pirate Library Mirror – a full catalogue of illegally acquired books – with other Anthropic employees. “just in time!!!” he wrote.
The legal reckoning
Project Panama only became public through thousands of pages of court documents unsealed in a copyright lawsuit brought by a group of authors, who alleged Anthropic illegally used their work to train Claude.
In 2024, a group of authors, led by novelist Andrea Bartz and nonfiction writers Charles Graeber and Kirk Wallace Johnson, hit Anthropic with a class action lawsuit, alleging the company violated copyright laws by using pirated books to train its AI model.
Last month, Anthropic – currently valued at US$965 billion (€875 billion) – agreed to pay $1.5 billion (about €1.3 billion) to settle the case, covering about 500,000 eligible works.
“It is the largest known copyright recovery in history,” said Justin Nelson, the authors’ lead attorney.
The case is the first of a wave of them against AI companies to settle.
Meta, OpenAI, Google and Microsoft are also facing copyright lawsuits from authors making similar allegations of piracy. Several authors who opted out of the class action settlement have individual cases pending against Anthropic, too.
What the ruling means
As other cases work their way through the courts, the settlement last week may be only the first chapter in a much longer legal fight.
In the original ruling last year, a judge found that the company’s use of lawfully purchased books was protected as fair use under US law.
However, the ruling drew a distinction between books the company bought and scanned and the millions of books it had previously pirated.
“The court’s landmark June 2025 ruling remains intact,” Anthropic deputy general counsel Aparna Sridhar said in a statement to the Post. “The issue we settled on was about how some materials were acquired, not whether we could use them to develop [AI].”
For some, however, the unsealed documents may have delivered the more lasting verdict, shattering whatever illusion may have remained that the world’s biggest tech firms understand cultural value. Let alone value the sanctity of human creation.