Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Anthropic has won a major victory in an ongoing legal battle on artificial intelligence models and copyright, which can resonate through the dozens of Other legal law prosecution Winding through the legal system in the United States. A court determined that it was legal for anthropic to train its AI tools on works protected by copyright, arguing that behavior is protected by the doctrine of “fair use”, which allows unauthorized use of documents protected by copyright under certain conditions.
“The use of training was fair use”, the senior district judge William Alsup wrote In a summary judgment released on Monday evening. In copyright law, one of the main ways of which the courts determine whether the use of works protected by copyright is fair use is to examine whether the use was “transformative”, which means that it does not replace the original work but rather something new. “The technology in question has been among the most transformative that many of us will see in our life,” wrote Alsup.
“This is the first major decision in a case of copyright generator of the AI to resolve fair use in detail,” explains Chris Mammen, director of Womble Bond Dickinson who focuses on intellectual property law. “The Alsup judge noted that the formation of an LLM is a transformative use, even when there is a significant memorization. He specifically rejected the argument according to which what humans do during reading and memorization are different in kind of what computers do when training an LLM.”
The case, a collective appeal brought by books of books who allegedly alleged that Anthropic had violated their copyright using their works without authorization, was deposited for the first time in August 2024 before the American district court of the North District of California.
Anthropic is the first artificial intelligence company to win this type of battle, but victory is fought with a large attached asterisk. While Alsup noted that anthropic training was fair use, he judged that the authors could take anthropic to test on their works.
While Anthropic finally went to training on copies purchased books, he had nevertheless collected and maintained a huge library of hacked materials. “Anthropic has downloaded more than seven million pirated copies of books, paid nothing and kept these hacked copies in her library even after having decided that she would not use them to form her AI (all or never). The authors argue that Anthropic should have paid for these pirated library copies. This order agrees, ”writes Alshropic.
“We will have a test on the hacked copies used to create the central library of Anthropic and the resulting damage,” concludes the prescription.
Anthropic did not immediately respond to requests for comments. The complainants’ lawyers refused to comment.
The trial, Bartz c. Anthropic was deposited for the first time less than a year ago; Anthropic asked for a summary judgment on the question of fair use in February. It should be noted that ILSUP has much more experience with the matters of fair use than the average federal judge, because he presided over the initial trial in Google c. OracleA historic case on technology and copyright which finally brought before the Supreme Court.
