Generative artificial intelligence is at the center of legal debates regarding its reliance on third-party copyrighted works for training data. Developers argue that such data is indispensable for achieving technological breakthroughs, as public domain and licensed data alone fall short. Central to these disputes is the concept of fair use, which developers claim transforms copyrighted materials into new, unrecognizable forms within AI models. This argument is contested in multiple legal and legislative arenas, with two federal cases—Thomson Reuters v. Ross Intelligence and Concord Music Group v. Anthropic—poised to shape the future of generative AI copyright law. Courts are set to rule on motions addressing unauthorized use of training data and the boundaries of fair use, potentially creating legal precedents that could guide future litigation.
Beyond courtroom battles, the U.S. Copyright Office is expected to deliver crucial guidance by the end of 2024 on the fair use of AI and the copyrightability of AI-generated works. Simultaneously, state and international regulations, such as the Colorado AI Act and the EU AI Act, are introducing transparency requirements for training data. In response, developers are proactively pursuing licensing deals and implementing safeguards to reduce output-side infringement risks. These efforts aim to lessen generative AI’s dependence on fair use arguments and mitigate legal challenges, signaling a pivotal moment for the technology’s future.
