Abstract: Major accidents in underground gold mines occur from time to time, causing serious casualties and property losses. In order to accurately evaluate the safety risk of underground gold mine, this paper puts forward a risk assessment model based on the combination of game theory and cloud model theory. Firstly, an assessment index system including 4 first-class indexes, including personnel factors, equipment factors, underground environment and safety management, and 28 second-class indexes is established, and the index risk levels are divided. Then, the weight of the indexes obtained by Decision Lab Analysis（DEMATEL） and Best-Worst Method（BWM） is combined optimally by game theory, and the importance and causality of the indexes are analyzed according to the centrality and cause degree obtained by Decision Lab Analysis（DEMATEL）. Finally, according to the cloud model theory, the standard cloud parameters of each risk level are determined, and the comprehensive cloud parameters are calculated and the assessment comprehensive cloud is generated. The model is applied to an underground gold mine in Shandong Province for safety risk assessment, and compared with fuzzy comprehensive assessment method for verification. The results show that the current safety risk level of the gold mine is lower（level II）, which is consistent with the actual assessment results of the project. The main factors affecting safety production are post length of service, educational level, safety awareness, monitoring equipment and lighting conditions. The model can solve the problems of relevance, fuzziness and randomness of assessment indexes, and can improve the accuracy of safety risk assessment results of underground gold mine.