CRiskEval is a Chinese dataset meticulously designed for gauging the risk proclivities inherent in LLMs such as resource acquisition and malicious coordination. The dataset consists of 14,888 questions simulating scenarios about 7 types of risk, each accompanied by 4 options which are generated by AI to describe different opinions or behavioral tendencies. The risk levels corresponding to the content of options are extremely hazardous, moderately hazardous, neutral and safe, and options are revised by humans to ensure that the content matches the risk level.