clearly indicated as such; and 3) when deepfakes are created using generative AI, this fact be disclosed or indicated in a manner that allows users to clearly recognize it. In the case of artistic or creative works, such disclosure may be made in a way that does not interfere with exhibition or appreciation. Under the AI Framework Act of Korea, the subjects of these obligations are AI business operators. Accordingly, civil society organizations that use generative AI tools may not themselves be the direct subjects of these legal obligations. However, given that the legal framework is still in a formative stage— with interpretations remaining fluid and amendments likely—and considering the underlying purpose of transparency obligations, it would be desirable for civil society organizations that prioritize trust and human rights to voluntarily uphold the principle of transparency to the greatest extent possible. However, requiring that the use of generative AI be uniformly disclosed on all outputs is unrealistic and may impose unnecessary burdens. As AI functions are increasingly built into internet search engines and office applications by default, situations are emerging in which AI is used—often to varying degrees—across a wide range of tasks regardless of the user’s intent. In such contexts, labeling every output with a statement such as “This output was created with the assistance of AI” would not only create practical burdens for organizations but also fail to provide meaningful information to audiences. Moreover, mechanically disclosing the use of AI for outputs that have been thoroughly reviewed and responsibly published by an organization may, paradoxically, undermine public trust in those 60 61