Abstract: With the steady development of new industrialization, the application of artificial intelligence (AI) technology in different stages of emergency management and different disaster situations is increasing, and its risks present new features and trends. However, there is still a lack of systematic study of the application of AI empowered emergency management and various types of risks. In order to improve the efficiency and effect of AI technology empowering emergency management, it is urgent to clarify the specific applications of AI in emergency management and its risk categories. Therefore, in combination with relevant literature and specific cases, the paper reviews the applications and potential risks of AI technology in emergency management. The review finds that AI technology is widely used in four periods: risk management, emergency preparedness, emergency response, and emergency recovery, covering 16 categories of specific applications such as risk early warning, simulation and prediction, psychological counseling, etc., which are applicable to a variety of scenarios such as floods and earthquakes. At the same time, based on the three dimensions of “human-AI-human AI interaction”, the paper summarizes nine types of risks and challenges faced by AI technology applied to emergency management, such as personal privacy and data security, data bias and algorithmic discrimination, and the degree of robustness of interaction, etc., and classifies the systemic risks of AI application in risk management in terms of the systems and algorithms, values and ethics, public acceptability, organizational acceptability and other factors from the dimension of human-AI interaction.