In a tale that echoes Ray Bradbury's dystopian visions, newly unsealed court documents have exposed how Anthropic, the company behind the AI chatbot Claude, ran a covert operation to scan and destroy millions of books. The project, aptly named Project Panama, involved slicing book spines with industrial machinery, scanning pages in bulk, and then pulping the paper for recycling. The digitized content was used to train Claude, aiming to improve its writing quality beyond what the internet could offer.
The secretive operation
Project Panama began in early 2024, according to internal documents. Anthropic executives sought books that predated online content, believing they would teach Claude "how to write well" rather than mimic "low quality internet speak." The company purchased tens of thousands of books at a time from vendors like Better World Books and UK-based World of Books. The heavy lifting—cutting spines and scanning pages—was outsourced to a vendor using hydraulic machines and industrial imaging equipment. The scale was staggering: the vendor was asked to convert between 500,000 and two million books over six months.
The internal planning document was explicit about secrecy: "We don't want it to be known that we are working on this." This revelation, first reported by the Washington Post, has cast a shadow over the AI industry's race for data.
Why books?
Anthropic's rationale was rooted in quality. Unlike the modern web, which is increasingly polluted by AI-generated content, books offer vast amounts of human-written, professionally edited text. This helps avoid "model collapse," a degradation that occurs when AI trains on its own output. But the method of acquisition raised legal and ethical red flags.
Court filings suggest that Anthropic executives may have believed they could get away with it. In June 2021, co-founder Ben Mann downloaded a trove of copyrighted content from LibGen, a "shadow library." A screenshot showed his browser in the act. The following year, he shared a link to the Pirate Library Mirror, a catalogue of illegally acquired books, with colleagues, writing "just in time!!!"
Legal reckoning
Project Panama came to light through a class-action lawsuit filed by authors including novelist Andrea Bartz and nonfiction writers Charles Graeber and Kirk Wallace Johnson. They alleged that Anthropic used pirated books to train Claude without permission. Last month, Anthropic—valued at US$965 billion (€875 billion)—agreed to pay $1.5 billion (about €1.3 billion) to settle the case, covering roughly 500,000 eligible works. "It is the largest known copyright recovery in history," said Justin Nelson, the authors' lead attorney.
The settlement is the first of its kind against an AI company, but not the last. Meta, OpenAI, Google, and Microsoft face similar lawsuits. Several authors who opted out of the class action have individual cases pending against Anthropic.
What it means for Europe
While the case is US-centric, its implications resonate across Europe. The EU has been at the forefront of AI regulation, with the AI Act setting global standards. The revelation of Project Panama could fuel debates in Brussels and national capitals about the ethics of AI training data. European publishers and authors, many of whom have seen their works used without consent, are watching closely. The settlement may set a precedent for how AI companies compensate creators, a issue that is also being litigated in European courts.
Moreover, the case highlights the broader tension between technological progress and intellectual property rights. As competitors like Alibaba's Qwen3.8-Max challenge Claude's capabilities, the pressure to acquire high-quality data will only intensify. The European Union's enlargement push and its focus on digital sovereignty may also influence how such disputes are resolved.
For now, the settlement leaves key questions unanswered. A judge's earlier ruling found that Anthropic's use of lawfully purchased books was fair use, but drew a line at pirated content. Anthropic's deputy general counsel, Aparna Sridhar, stated: "The court's landmark June 2025 ruling remains intact. The issue we settled on was about how some materials were acquired, not whether we could use them to develop [AI]."
As the legal battles continue, the story of Project Panama serves as a cautionary tale about the lengths companies will go to feed their algorithms. It also underscores the need for transparent and ethical practices in the AI industry, a concern that is as relevant in Berlin and Paris as it is in San Francisco.
