Anthropic’s $1.5 billion copyright settlement with authors is being cast as a landmark win for creative workers. Half a million authors stand to receive $3,000 each, and the payment is the largest recovery in the history of U.S. copyright law. It looks like a victory lap for the Authors Guild, the Association of American Publishers, and the plaintiffs who pressed the case. But once the dust settles, the real beneficiary may be the AI industry itself.
The key legal question wasn’t whether Anthropic pirated books from Library Genesis and Pirate Library Mirror: it did, and Judge William Alsup called the conduct “inherently, irredeemably infringing.” The question was whether training a model on copyrighted material, once lawfully acquired, is permissible. On that, the judge was clear: it is. In June, Alsup wrote that Anthropic’s training use “was a fair use,” describing it as among the most transformative technologies of our lifetimes.
That ruling survived the settlement. Anthropic will pay out for past misconduct, delete pirated datasets, and move on. But the broader precedent, that using copyrighted works for training can be fair use, remains intact. In fact, it has been strengthened.
