생성형 AI의 저작권 침해에 대한 검토-일본에서의 논의를 중심으로-

Abstract

There are two cases of copyright infringement in relation to production AI. One is the infringement of the right to copy and transmit due to collecting and processing original works during the development and learning stage of AI, and the other is the infringement of copyright due to the results generated using AI. The goal of this paper is to explore the infringement issues and solutions for these two cases. First, regarding the infringement of the right to copy and transmit due to collecting and processing original works during the development and learning stage of production AI, in the case of works used for learning by production AI, rather than infringing the ‘right to copy’ as generally assumed by the Copyright Act, it is intended to ‘learn’ like the function that occurs in our brain. However, since this is done through information processing devices such as computers, ‘learning’ is done through ‘copying’ or ‘transferring’, so it takes the form of infringement of the right to copy and transmit in the traditional sense. However, in the case of production AI, it is difficult to assume actual infringement of the copyright holder’s copyright through copying or transmission alone during the learning stage. Even if it is actually copied, it is difficult to conclude that it is a copyright infringement unless it is published. In addition, if the output from the generative AI is substantially similar to an existing work, it can be dealt with as a copyright infringement issue in the generation and use stages described later, and it is unlikely that it will be necessary to view it as a copyright infringement from the learning stage. Next is the copyright infringement according to the results generated using AI, and this always involves the issue of reliance. Regarding ‘reliance,’ the most problematic case is when the generative AI user is not aware of the existing work, but the work is included in the AI ​ learning data. In this case, if the result is produced using the generative AI that learned the work, it can be considered similar to unknowingly encountering the work because the AI ​was used as a tool. Therefore, in this case, it should be determined that there is reliance.