A copyright battle that has captured the entire AI industry has finally reached a verdict. A report from Associated Press in San Francisco shows that a federal judge has approved a $1.5 billion copyright settlement agreement for Anthropic: the artificial intelligence company will pay approximately $3,000 per book to thousands of authors for using unauthorized copies of their works to train its Claude chatbot.

In Monday's ruling, the presiding district court judge Araceli Martínez-Olguín stated that this class-action settlement provides meaningful relief to affected authors and publishers. The distribution of the money is now clear — about 91% of the more than 482,000 books covered by the ruling have already been claimed by authors or publishers, who are now in line for compensation.

Justin Nelson, lead attorney for the plaintiffs, said in a statement that this settlement is the largest known copyright recovery in history and expressed anticipation for the swift distribution to collective members. The weight of this statement lies in the fact that it brings the scale of compensation in an AI training data infringement case to the forefront of copyright history.

Looking back at the path that led to the ruling reveals the legal reasoning behind this settlement. In September last year, U.S. District Court Judge William Alsup in San Francisco gave preliminary approval, after which he retired. Alsup had previously issued a mixed ruling in the summer of last year: on one hand, he determined that training AI chatbots with copyrighted books was not illegal; on the other hand, he ruled that Anthropic obtained millions of books through unauthorized websites, which was itself unjustifiable. The distinction between lawful training and unlawful acquisition became the most critical legal pivot in the case.

Anthropic's response to the ruling focused on confirming the legality of the training. Aparna Sridhar, the company's Deputy General Counsel, emphasized that the ruling is a milestone, indicating that training AI with books falls under fair use under copyright law. For Anthropic and the entire large model industry, this statement is worth more than the $1.5 billion price tag — it means that the judiciary has given a favorable endorsement to the training activities that have already taken place.

When the $1.5 billion settlement is approved by the judge and the $3,000 per book compensation is about to be distributed, the real legacy left by this lawsuit is not the check itself, but two lines it has drawn: obtaining data through piracy is costly and unjustifiable, while training models through legal means is gradually recognized as fair use by the judiciary. The space between these two lines is where every AI company must tread carefully in the future.