Anthropic, the AI startup, can now start paying authors and book publishers after a major legal battle over copyright infringement. On Monday, a federal judge gave the green light to Anthropic’s huge $1.5 billion settlement agreement in a class action lawsuit, according to Reuters.
The case dates back to when Anthropic was accused of illegally downloading and storing millions of copyrighted books to train its AI models. Last year, U.S. District Judge William Alsup from the Northern District of California gave preliminary approval to the settlement after ruling that Anthropic had indeed obtained many books unlawfully.
Since then, Judge Alsup retired, and the final approval came from Judge Araceli Martinez-Olguin this Monday, officially closing the chapter on this lawsuit.
How the Settlement Works
Under the terms of the settlement, Anthropic will pay $3,000 for each copyrighted work involved—estimated at around 500,000 books and other written materials. These payments will be distributed among the rights holders, including authors and publishers.
The $1.5 billion figure makes this settlement one of the largest in U.S. copyright history. But despite the size of the payout, many authors and creators aren’t exactly celebrating.
Why Some See This as a Mixed Result
The controversy stems from how the court handled the core legal question. Judge Alsup ruled that training an AI model on copyrighted text qualifies as “fair use,” a landmark decision that many see as a win for the AI industry. Essentially, it means AI companies can use copyrighted material to train their systems without the need for explicit permission.
However, the ruling also made clear that the way Anthropic acquired some of these books wasn’t legal. The company built its training dataset from two sources:
- Books it legally purchased and scanned.
- Books it downloaded from pirate sites like Library Genesis and Pirate Library Mirror.
While the court said training on copyrighted text is fair use, it didn’t excuse Anthropic’s use of pirated copies. Judge Alsup found that downloading from pirate sites violated copyright law and could potentially lead to a trial. To avoid that, Anthropic agreed to this massive settlement.
What This Means for the Broader AI and Copyright Debate
Even though this settlement wraps up the Anthropic case, it doesn’t settle the wider legal questions around using copyrighted works to train AI models. That’s because Judge Alsup’s ruling was from a single district court, and by settling, Anthropic avoided an appeals process that could have set binding national precedent.
Other courts and judges are still free to rule differently on similar cases, and that’s exactly what’s happening. Multiple copyright lawsuits against big tech and AI companies like Google, Meta, Midjourney, and OpenAI are still ongoing. These cases are testing whether training AI on copyrighted content without permission is lawful.
For example, just last week, a group of publishers and authors—including Hachette, Cengage, Elsevier, author Scott Turow, and the organization S.C.R.I.B.E.—filed a class action lawsuit against Google. They allege that Google used their copyrighted materials without authorization to train its AI platform, Gemini.
Looking Ahead
The Anthropic settlement is a landmark moment in AI copyright law, both for its size and for the legal questions it highlights. While it provides some financial relief to affected authors and publishers, the industry is far from reaching consensus on what’s fair and legal when it comes to training AI on copyrighted content.
As the technology evolves and more lawsuits come through the courts, we’ll likely see more definitive answers emerge—though it may take years before there’s any real clarity.