In the early stages of the development of autonomous driving technology, autonomous vehicles (AVs) and manually driven vehicles (MVs) will both be present on the roads, and the interaction of AVs and MVs will affect driving safety. This study aims to evaluate driving safety for car-following scenarios involving AVs and MVs on urban roads and to identify the driving safety vulnerability sections for each road. In this study, the AV behavior control algorithm and the urban interrupted road were implemented using SCANeRTM, and longitudinal, lateral, and inter-vehicle driving safety indicators were derived. As a result of the analysis, the driving safety of the AV-AV pair was the highest in all safety indicators, and the mixed pair (AV-MV) was safer than the MV-MV pair. The driving safety evaluation index for each analysis section was analyzed by the rate of change. As a result of analysis of variance, the null hypothesis that the section was the same in all mixed pairs of evaluation indicators was rejected. Post hoc analysis shows that the section with the greatest difference from the straight line was selected as a vulnerable area. As a result of the post hoc analysis, the non-signal intersection was analyzed as the most vulnerable area in the case of the mixed pair. Using this, it is possible to select a driving safety vulnerability section when AVs and MVs are mixed on actual urban roads.