Generative AI and Copyright Issues : Cases and Analysis of Copyright Infringement and Copyrightability

With the rapid advancement of generative AI, anyone can easily produce creative works in fields such as literature, music, and visual art. In particular, deep learning-based generative AI systems that generate content by training on large volumes of existing copyrighted works pose a significant challenge to the existing copyright framework. Based on selected copyright litigation cases related to generative AI, this paper analyzes and examines two major legal issues under copyright law: (1) Copyright infringement during the training and use of AI models, and (2) The copyrightability of AI-generated outputs. In the AI training phase, the unauthorized reproduction of copyrighted works, removal or alteration of CMI, and unauthorized use can constitute copyright infringement. Furthermore, if the AI-generated output is substantially similar to the original works used in AI training, the use of such outputs may also result in infringement. Additionally, when AI platforms act as OSPs and induce or facilitate infringing behavior by users, questions of secondary liability may arise. Regarding copyrightability, it is important to distinguish between AI-generated outputs, which involve no human creative input, and AI-assisted outputs, which incorporate human creative contributions using AI as a tool. Both the U.S. Copyright Office and the Korea Copyright Commission have adopted this distinction as a basis for determining copyright registration eligibility in their respective practices. This study provides valuable insights into the potential application and interpretation of existing copyright law in the context of generative AI.