US, China Courts Issue Landmark AI Copyright Rulings

Recent court rulings in the U.S. have significantly shaped the legal landscape around generative AI and copyright, favoring AI companies in the use of copyrighted works for training large language models (LLMs). In cases involving Anthropic's Claude and Meta's LLaMA, courts found that scanning and ingesting copyrighted books for AI training constitute fair use, emphasizing the transformative nature of such use and the lack of harm to the market for original works. These rulings clarify that using copyrighted texts to train AI models is necessary and reasonable, focusing on learning language patterns rather than reproducing original content. However, the decisions also indicate that fair use defenses may not always hold, especially for content types other than books, leaving some uncertainty for rights holders. Meanwhile, in China, a landmark ruling held an AI platform liable for copyright infringement for facilitating the generation and dissemination of AI-created images resembling protected characters, rejecting the platform's safe harbor defense. Together, these cases highlight a global legal tension between protecting IP rights and fostering AI innovation, with courts beginning to define the boundaries of copyright in the era of generative AI.