US$1.5 billion settlement settled: The judge approved the Anthropic pirated book case, with a compensation of US$3,000 per book, marking a key precedent for AI training copyright. A copyright tug-of-war that affected the entire AI industry finally came to the judge's hand.

US$1.5 billion settlement settled: The judge approved the Anthropic pirated book case, with a compensation of US$3,000 per book, marking a key precedent for AI training copyright. A copyright tug-of-war that affected the entire AI industry finally came to the judge's hand. An Associated Press report from San Francisco shows that a federal judge has approved Anthropic’s copyright settlement of up to $1.5 billion: The artificial intelligence company will pay thousands of authors about $3,000 per book for using pirated copies of the authors’ works to train its Claude chatbot. District Court Judge Araceli Martínez-Olguín, who presided over the case, said in Monday’s ruling that the class-action settlement provides meaningful relief to affected authors and publishers. The whereabouts of the money already have a clear outline - about 91% of the more than 482,000 books covered by the ruling have been claimed by authors or publishers, and they are now in the queue for compensation. Justin Nelson, an attorney for the plaintiffs, said in a statement that the settlement is the largest known copyright recovery in history and said he looks forward to completing distribution to class members as soon as possible. The weight of this sentence lies in the fact that it puts the scale of compensation in AI training data infringement cases at the forefront of copyright history for the first time. Turning back time to the origin of the ruling, we can clearly see the legal context behind this settlement. U.S. District Judge William Alsup, who has since retired, gave preliminary approval last September in federal court in San Francisco. Alsup issued a mixed ruling last summer: on the one hand, it held that it was not illegal to train an AI chatbot with copyrighted books; on the other hand, it ruled that Anthropic mistakenly obtained millions of books through pirated websites, which itself was untenable. The dividing line between legal training and illegal acquisition has become the most critical legal fulcrum of the entire 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 this ruling is a milestone and shows that using books to train AI is fair use under copyright law. For Anthropic and even the entire large model industry, this sentence is more valuable than the price tag of US$1.5 billion - it means that the judiciary has given a tendentious endorsement to the training behavior itself that has occurred. When the $1.5 billion settlement was approved by the judge and the compensation of about $3,000 per book was about to be paid out, the real legacy left by this lawsuit was not the check, but the two lines it drew: using piracy to obtain data, which is costly and untenable; using legal means to train