search). ② Users are encouraged to substantially modify or edit generative AI outputs before using them. Generative AI systems rely on a wide range of data for training, including publicly available data as well as data obtained through separate contractual arrangements. Such data may include not only personal data but also copyrighted works. These works encompass various formats, including literary works such as poetry and novels, music, images such as photographs and illustrations, and audiovisual works. Some works are no longer protected because their copyright term has expired. Under Korean copyright law, economic rights are protected for 70 years after the author’s death, and for works made for hire, for 70 years after publication. Conflicts between AI companies and copyright holders over the use of copyrighted works for AI training have become a highly contentious global issue, with numerous lawsuits currently underway. In some cases, individual licensing agreements are concluded between AI companies and rights holders, but as of November 2025, these issues remain far from being conclusively resolved. There are diverging views on this matter, including arguments that copyright should be strictly protected and counterarguments that the use of works for AI training should be permitted as fair use. A detailed discussion of these debates, however, falls outside the scope of this guide. However, civil society organizations, as users of generative AI, may themselves become involved in copyright disputes in the course 86 87