A Study on Copyright Issues on Music Works used in Generative AI

This study deals with a variety of issues related to music copyrights related to Generative AI products. The main issue is whether copyright infringement is established or not when music works are reproduced and transformatively used with Generative AI learning data without any permission of copyright owner. However, most of the music works used in the Generative AI learning stage are transformatively used, and at this time, it is difficult for the music copyright holder to obtain information unless the Generative AI platform operator discloses or informs how his work was used in the Generative AI learning stage. Therefore, even if a music copyright holder files a copyright infringement lawsuit against an Generative AI developer or platform operator, there are difficulties due to the problem of the burden of proof, and in this regard, it will be necessary to consider ways to apply the relaxation of the burden of proof, such as medical lawsuits or industrial accident lawsuits. Since Generative AI products do not meet the requirements as human authorship, it is difficult to be protected as works under the current copyright law, so a “Generative AI register” only for Generative AI products should be created and operated by the Korea Copyright Commission. In other words, it is possible to consider dualizing the registration system so that AI products can be registered separately from human works, and to register the name of the person who made the Generative AI product, the title of the product, etc. For Generative AI products, it will be possible to shorten the protection period to 3 to 5 years which should be applied according to the basis principle of calculating from the time of publication. In addition, when music works are used as Generative AI learning data to generate AI music output, it is necessary to make a system that can conclude contracts between Generative AI developers and music copyright owners or pay compensation systematically. In particular, it is necessary to consider the introduction of an Opt-Out method that allows Generative AI developers to create music by considering it as agreed unless the original music copyright holder explicitly rejects it. In addition, it is necessary to seek legislative solutions through further discussions on Generative AI plagiarism, mandatory display of AI music products, sound source sampling, and legal responsibility as an Online Service Provider (OSP) of Generative AI platform operators. There are currently no international norms regarding the legal protection of works used without any permission in Generative AI learning data, so different legal standards are applied to each country, and it is expected that international discussions will take place in the future to establish guidelines or international treaties.