The arrival of Level 2 (L2) and Level 3 (L3) of automated driving places new demands on driving behavior, yet drivers who have been accustomed to manual driving do not seem to be adequately prepared. Therefore, to improve driving safety, this study proposes an integrated approach to establishing a safe driving behavior model by combining Hierarchical Task Analysis and Information Decision Action model. Initially, 11 key driving scenarios at the L2 and L3 levels were identified through literature review; Five automated driving experts were invited to grade the importance of the scenarios through the Analytic Hierarchy Process. Additionally, their safety recommendations for each scenario at the L2 and L3 levels were analyzed. Finally, a model of drivers’ safe behavior under L2 and L3 conditions was constructed accordingly. The results of this study can not only guide the cultivation of drivers’ driving habits, but also provide test scenarios with human-machine co-driving perspectives for the development of automated driving.