created content or for real-world events, as is often the case with deepfakes. One such example occurred when a fake image depicting what appeared to be a large explosion near the U.S. Department of Defense headquarters (the Pentagon) spread on social media, causing widespread confusion. In some cases, transparency may be a legal obligation. For example, under the EU AI Act 1) providers of AI systems must design their systems so that people are aware when they are interacting with an AI system; 2) in the case of generative AI, providers must ensure that outputs can be recognized as AI-generated content in a machinereadable manner. In addition, 3) deployers (users) of emotion recognition or biometric identification systems must inform natural persons that such systems are being used; and 4) deployers (users) of AI systems that generate deepfakes must disclose that the output has been generated by AI. In the case of artistic works, however, this disclosure may be made in a way that does not interfere with the appreciation of the work. For civil society organizations, the fourth obligation is likely to be particularly relevant, as such organizations often create parody images criticizing power or produce documentaries related to specific social issues. Korea’s AI Framework Act also establishes obligations to ensure AI transparency (Article 31). Specifically, it requires that: 1) users be informed in advance when a service is operated based on highimpact AI or generative AI; 2) outputs generated by generative AI be Generative AI Guide for Civil Society