Federal judge Araceli Martinez-Olguin definitively approved on Monday a $1.5 billion settlement between Anthropic and a group of authors and publishers who sued the company for training its AI model Claude with protected books without authorization. With this ruling, the company closes one of the most significant copyright litigation cases in the artificial intelligence era.

How Anthropic ended up in court

In 2024, a collective of writers sued Anthropic after discovering that the company had used copies of their works to train Claude without consent. The case took an unexpected turn in June 2025, when judge William Alsup ruled that training AI models with copyright-protected books falls within "fair use"—a decision that became precedent for the entire industry.

However, the ruling also recognized that Anthropic had committed specific infractions: downloading more than 7 million copies from digital portals without authorization and storing them in its own internal library. That specific point would have led to a separate trial with potentially much higher damages, so the company chose to negotiate.

The 'Panama' project: from LibGen to bookstores

Court documents revealed the real scope of the text acquisition program, as detailed by xataka.com. Anthropic not only downloaded books from repositories like LibGen: the company even physically purchased and scanned millions of volumes, cutting their spines to digitize them and recycling them afterward.

Company employees, including co-founder Ben Mann, directly participated in these unauthorized downloads. The company has always maintained that this material was never used to train commercial models, although that argument failed to prevent litigation.

Distribution: $3,000 per work and over 91% claimed

The settlement establishes compensation of approximately $3,000 per work, based on an estimated total of 500,000 titles among authors and publishers. According to TechCrunch, over 91% of those affected have already requested their share of the payment.

Plaintiffs' attorneys will receive $101 million in fees—below the $187.5 million they had initially requested—a figure the judge considered proportionate to the risks that continuing litigation would have entailed.

Criticism and objections that did not prevent approval

Some authors filed objections arguing that the amount was insufficient, that attorney fees were excessive, or that the settlement unjustly excluded certain rights holders. Judge Martinez-Olguin rejected these arguments, deeming them unrealistic regarding what continuing with a risky trial would have meant.

Some authors and publishers have chosen not to participate in the settlement and are maintaining separate lawsuits against Anthropic, so litigation could continue in court for specific cases.

What this means for AI and copyright

This settlement marks an important milestone: Justin Nelson, lead attorney for plaintiffs, called it "historic" and stated that it represents the largest copyright recovery ever achieved. However, the message is nuanced: while training AI with protected works is considered legitimate, companies cannot make massive, uncontrolled stockpiles of content without consent.

For other technology companies developing language models, this ruling establishes clear boundaries on how they can obtain training data, although the legal battle over AI and copyright is far from over.