Intelligent transportation systems have significant potential to improve road safety. Given that traffic parameters can be related to crash risk, and traffic parameters on road networks are complex and dynamic, it is problematic to use static crash risk metrics based on historical crash data to quantify safety. While conflict-based techniques have been used to establish more proactive safety metrics, Bayesian hierarchical extreme value models allow for the quantification of crash risk based on conflicts as they appear in real time. Previous studies using drone data have shown that crash risk derived from these models change dynamically with time and vary by site. Since drone data is continuous, full coverage of the network is available. Drone data, however, presents several scaling challenges. Autonomous vehicles may be used to obtain data on the vehicle’s surroundings, treating it as a mobile sensor. This study explores the applications of real-time crash risk using autonomous vehicle data, addressing issues around the use of a moving observer, temporally discontinuous data, and the underrepresentation of specific sites. A dataset from Boston was used to quantify network-wide crash risk. Crash risk was found to vary both spatially and temporally. Moreover, the chronicity of crash risk in each location in the network was also identified, with the locations exhibiting the highest crash risk being locations with low chronicity. These insights may inform ITS solutions to address short-term peaks in crash risk, including crash risk monitoring dashboards, routing applications, and adaptive traffic signal control systems.