The ink is dry. A federal judge in San Francisco has officially stamped her approval on a $1.5 billion settlement. This is the largest copyright class action settlement the history books will record.
Anthropic is paying. Authors are getting cash. The legal fight over how the AI giant built its knowledge base is finally, conclusively, over.
How the Case Was Settled
U.S. District Judge Araceli Martínez-Elguín signed the final order on July 20. The case was brought by lead plaintiffs like Andrea Bartz and Kirk Wallace Johnson. They argued that Anthropic stole books.
Specifically, they pointed to pirated libraries. LibGen and PiLiMi. Anthropic used these shadow archives to build its book collection.
But here is the twist that makes this settlement unique. The lawsuit never argued that training AI on copyrighted material is illegal. The court had already ruled that training itself is fair use.
The underlying dispute centered entirely on how Anthropic acquired the books.
The issue wasn’t the model. It was the method. Illegal acquisition. Plain and simple.
Who Gets the Money
So, who benefits from this half-billion-plus pile?
Authors and publishers whose works landed on Anthropic’s “Works List.” The compensation? Roughly $3,000 per infringed book.
Think about that number.
That is about four times the usual statutory minimum for copyright infringement cases in the U.S. It is a significant payout for creators whose work was scraped without permission.
The uptake has been massive. Over 440,00 books have been claimed. That accounts for more than 91% of eligible works. Most authors have taken their cut.
What This Doesn’t Cover
Don’t get it twisted. This settlement is narrow.
It only covers past liability. How the data was gotten in the past. That ship has sailed.
It does not shield Anthropic from future lawsuits. If an author feels their work caused harm through the chatbot’s output, they can still sue. The judge was explicit. The release does not cover future harm. It doesn’t touch claims “based on the output of AI models.”
The door remains open for new claims. Going forward, everything is fair game.
The Rejected Demands
It wasn’t a free-for-all. The court heard from 54 different objections and comments. Class members and third parties had demands.
Some wanted the list of covered works expanded. Others wanted non-monetary fixes. Source attribution. A big ask. Then there were the hardliners who wanted Anthropic to delete its models entirely.
Judge Martínez-Elguín overruled them all.
Her reasoning? Those requests went beyond what this specific lawsuit could address. This was about compensation for theft, not a referendum on the existence of AI itself. The court kept its scope tight.
The Aftermath
With the order finalized, the case status changes from “active” to “closed.”
But the job isn’t entirely done. The court is keeping an eye on the money distribution. Ensuring the authors actually get what they were promised.
We know the precedent. We know the payout. But we don’t know what Anthropic does next. They have their data. They have their settlement. They still have a chatbot that generates text based on that same data. The output remains the variable no one can fully control.
