In a landmark resolution, Anthropic, the company behind the AI chatbot Claude, has agreed to compensate authors whose books were used without permission to train its language models. Under the settlement, each eligible work could earn its rights holder up to $3,000, with the total payout capped at approximately $1.5 billion plus interest. The agreement, announced by the plaintiffs, covers around 500,000 books that were allegedly sourced from the pirated LibGen database.
This settlement marks one of the first major financial reckonings for the AI industry over the unauthorized use of copyrighted material. For many writers, the payout offers a rare financial reprieve in a profession where advances are often modest and income has dwindled as reading habits shift and AI tools increasingly absorb creative work. While the amount per book may not match the profits tech giants reap from AI, it represents a tangible acknowledgment that authors' labor has value.
Notably, the settlement does not grant Anthropic blanket immunity. The agreement explicitly allows individual authors to pursue separate legal action if they believe their work is still being used illegally. Additionally, Anthropic has committed to deleting the pirated copies it downloaded, a step that addresses a core grievance of the plaintiffs.
Why the Settlement Matters
Legal experts have drawn parallels to the downfall of Napster, the early file-sharing service that collapsed under a wave of lawsuits from musicians. As one lawyer told The New York Times, this case could be "the AI industry's Napster moment," signaling a shift in how courts and companies treat the unlicensed use of creative works in training data.
For authors, the practical question is how to claim their share. Writers who believe their books were used can visit the settlement website to file a claim. To verify eligibility, they can search the LibGen database, which was the source of the allegedly pirated texts. The process is straightforward, but the deadline and documentation requirements have not been fully detailed.
One author who stands to benefit is Kyle Chayka, a staff writer at The New Yorker. His books "Filterworld: How Algorithms Flattened Culture" (2024) and "The Longing For Less: Living With Minimalism" (2020) were found in LibGen, along with an Italian translation of "Filterworld." If all three are covered, Chayka could receive upwards of $12,000.
In an interview, Chayka reflected on the settlement's terms. "It should be a license, really," he said, "because the training never goes away. So it could be $5,000 every 5 years, or $1,000 / year as long as they exist. But the price seems about right, honestly — a decent percentage of most book advances, and about the price of an institutional speaking gig."
Chayka also downplayed the idea that his books would accelerate AI's environmental and societal harms. "Even if AI couldn't be trained on any book ever produced, tech would still find ways around it; the real licensing money will be 'new data' aka fresh journalism," he wrote. "So outside of my preexisting enmity and hatred for the way that AI is destroying civilization and the planet, I don't think the books make it much worse."
When asked what he might do with his potential payout, Chayka joked about buying a stone townhouse in an Italian village without wifi — a lighthearted note in an otherwise serious legal development.
As the settlement moves toward final approval, it could set a precedent for how other AI companies handle the use of copyrighted works. The outcome may influence ongoing negotiations and lawsuits involving other tech firms, and it underscores the growing tension between innovation and intellectual property rights.
