생성형 AI에 의한 창작물과 저작권

AI(Artificial Intelligence) is rapidly transforming our economy and society. Over the past few years, AI has become a fundamental part of our daily lives, introducing revolutionary products and services into our homes and workplaces. Consequently, it enables various tasks and activities, creating numerous new opportunities that were unimaginable just a decade ago. Many people anticipate that computers and AI will assist in the production of useful products. Recently, generative AI systems like ChatGPT, BARD, DAL-e, Midjourney, and Stable Diffusion have emerged, allowing anyone to utilize AI for the creation of literature, plays, and artworks. With the emergence of generative AI systems, questions have arisen about whether AI-generated works can meet copyright protection requirements, and if so, the scope and ownership of such protection. Furthermore, concerns arise about potential copyright infringement when AI-generated creations resemble existing works. The use of third-party works as training data for developing generative AI is also a subject of interest. Many countries, including South Korea, are grappling with the issue that the ownership of creative works generated by generative AI, remains in most cases, where copyright requires a human creator. Copyright can be recognized when AI creates imaginative and unpredictable works that encompass new ideas or expressions. Generally, the copyright of works created by AI belongs to the developers, companies, or individuals who developed the AI. Consequently, the copyright of content generated by AI falls to these developers or companies. However, as recognized in some countries, there's a need to consider whether generative AI should be acknowledged as a joint copyright holder. When AI generates a work, it can be based on original works or existing creations. Countries including Japan, the EU, the UK, and the US have established special provisions in copyright laws to address copyright issues related to data training and processing in artificial intelligence technologies. They do so to provide exemptions or consider such activities under fair use principles without being seen as copyright infringement. The EU generally includes exceptions in copyright laws to allow the use of others' works freely for non-commercial and academic purposes in AI technology. Japan has clear provisions that allow the use of data within a copyrighted work without the copyright holder's consent, as long as the underlying ideas are not conveyed. In South Korea, legal uncertainties still exist regarding the issue of using copyrighted works as training data for AI. While fair use might allow such use, there's a lack of clear precedent regarding the legality of mechanical data collection and analysis under fair use. Even if the use of training data for AI falls under fair use, there's a risk of potential copyright infringement judgments in court. Therefore, to eliminate legal uncertainties surrounding these issues, specific legal regulations need to be established.