In a significant development for artificial intelligence (AI) and copyright law, a federal judge in California delivered a pivotal ruling on Monday. The decision, one of the first major court judgments in this area, addressed the legality of using copyrighted materials to train large language models (LLMs). The ruling deemed that training these AI systems on copyrighted books is permissible under the fair use doctrine, provided the materials are legally obtained.
The lawsuit, filed against Anthropic, a prominent AI company known for its AI platform “Claude,” was initiated by three authors who alleged the misuse of their copyrighted works. According to the lawsuit, Anthropic used both legally acquired and pirated books to train its systems. Judge William Alsup, writing for the US District Court for the Northern District of California, clarified that digitizing purchased books for AI training constitutes a mere format change and does not infringe copyright protections.
This court ruling hinges on the “transformative” nature of AI training, a key component of the fair use doctrine outlined in the Copyright Act of 1976. The Act specifies four factors to assess the legality of using copyrighted material: the purpose and character of use, the nature of the work, the portion used, and the market impact. The court’s decision aligns with the 1994 Supreme Court ruling that transformative use often constitutes fair use.
While the decision represents a success for AI firms, it also underscores the necessity of only using legally sourced materials. Anthropic was granted summary judgment regarding its utilization of purchased books, but accusations of using pirated content will proceed to trial.
As technological advancements continue to blur the lines between AI and copyright law, the landscape remains dynamic and complex. Legal practitioners and corporations will need to stay vigilant as courts further define these boundaries. More details on this ruling can be found at JURIST.