A U.S. court has approved a $US1.5 billion ($AU2.1 billion) settlement between AI company Anthropic and a group of authors and publishers.
Authors brought the case because Anthropic downloaded millions of pirated books to train its chatbot Claude.
Anthropic will now pay out about $US3,000 per pirated book to thousands of authors and publishers.
Some argue that’s not enough, given the scale of the infringement.
Editorial note: The Daily Aus has a commercial partnership with Anthropic on our newsletters and podcast. Partnering with a range of brands keeps our content free for you. Anthropic has no involvement or influence on our editorial content, including this piece.
Background
The case began when three authors – Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson – sued Anthropic in 2024.
They alleged the company downloaded millions of books from sites including Library Genesis and Pirate Library Mirror to train its AI models.
These sites illegally hosted copies of books so people could download and read them for free (called pirating).
The case became a class action, with Bartz, Graeber, and Johnson as the lead plaintiffs representing other publishers and authors.
Judgment
In 2025, Anthropic asked the judge in the case, William Alsup, for a summary judgment.
Under U.S. law, a summary judgment is a way of securing a legal finding without a trial when there’s no dispute over the central facts.
Anthropic wanted a judgment on the U.S. legal concept of “fair use”: copyrighted material can be used freely for research or education, or if it’s “transformed” in some way.
Alsup heard that Anthropic downloaded millions of pirated books to train Claude, then moved to buying paper books and digitising them for the same purpose.
Alsup ruled: “The use of the books at issue to train Claude and its precursors was exceedingly transformative and was a fair use... The digitisation of the books purchased in print form by Anthropic was also a fair use... However, Anthropic had no entitlement to use pirated copies for its central library.”
Your contribution ensures The Daily Aus can continue doing the work you love.
I want to give $10 monthly to help TDA keep doing the news.
Your details:
TL;DR: Pirating the books was wrong, but how Anthropic used the pirated books was fine.
Settlement
The finding left Anthropic facing a jury trial to work out what it owed, which could have been hundreds of billions of dollars if the maximum damages were applied across every pirated book.
Instead, Anthropic settled for $US1.5 billion, which works out to roughly $3,000 per book in the class action.
Many authors will split this with their publisher.
As part of the settlement, Anthropic must also destroy the pirated files it downloaded.
Response
CEO of the Authors’ Guild Mary Rasenberger said the settlement sends “a strong message to the AI industry that there are serious consequences when they pirate authors’ works to train their AI.”
Anthropic lawyer Aparna Sridhar told TDA the ruling showed “that training AI on books is fair use under copyright law.”
Some members of the class action argued the settlement was too small given the scale of the alleged piracy and the range of possible damages available under the Copyright Act.
What’s next?
This settlement only covers the pirated books Anthropic downloaded to train its models.
It says nothing about what those models produce, so more lawsuits could still be on the way.
It also doesn’t set a wider precedent, since the fair use ruling was never tested on appeal.
Other cases could still land on a different answer about whether AI can be trained on copyrighted works.
That includes cases from other authors who opted out of this class action to pursue their own separate claims against Anthropic.
