Anthropic has achieved a significant victory in a legal battle on ongoing AI models and copyrights that could reverberate in other AI copyright lawsuits in the United States. The court has determined that AI tools for humans to train copyrighted works are legal, and believes that the act is obscured by the “fair use” doctrine that allows unauthorized use of copyrighted materials under certain conditions.
“Training uses are a reasonable use,” Senior District Judge William Alsup wrote in a summary judgment order issued Monday night. In copyright law, one of the main ways courts determine that copyrighted works are used without permission is to check whether use is “transformative”, meaning it cannot replace the original work, but some new work. “The technology in question is one of the most transformative technologies in many of our lives,” Alsup wrote.
“This is the first major ruling in the generation of AI copyright cases to address fair use in detail,” said Chris Mammen, managing partner at Womble Bond Dickinson, who focuses on intellectual property law. “Judge Alsup found that training LLMs is transformative use – even when there is a significant memory. He specifically rejected the argument that humans do differently when reading and remembering than computers do when training LLMs.”
The case was a class action lawsuit filed by the author of the book, claiming that humans violated their copyrights by infringing on unauthorized works, filed in August 2024 in the U.S. District Court for the Northern District of California.
Anthropomorphism was the first AI company to win such a battle, but the victory was accompanied by a large asterisk. Although Alsup found that anthropomorphic training was justified, he ruled that the author could accept piracy by human opponents.
Although humans eventually moved to training on purchased copies of books, it still first collected and maintained a huge library of pirated materials. “Authorization downloaded over 7 million pirated books, gained nothing, and kept copies of these pirated copies in its library, even in the decision that it would not use them to train its AI (at all or again). The authors believe that anthropomorphism should pay for these pirated libraries. This order agrees,” Alsup wrote.
“We will experiment with the pirated copies used to create the Central Library of Humanity and the resulting losses,” the order concluded.
Humans did not immediately respond to requests for comment. The plaintiff’s attorney declined to comment.
litigation, Bartz v. Humans First submitted less than a year ago; humans demanded a summary judgment on the issue of fair use in February. It is worth noting that Alsup has experience with fair use issues than an ordinary federal judge, as he presided over the initial trial Google v OracleThis is a landmark case about technology and copyright that eventually emerged in the Supreme Court.
