The objective of this technical paper is to present a method that characterizes autonomous vehicle (AV) safety performance through the application of risk-based validation that leverages existing crash incidence and severity data, physics based model and simulation, and U.S. Federal Motor Vehicle Safety Standard (FMVSS) benchmark metrics. The output of the proposed risk-based methodology is a framework that organizes the number and type of physical tests and model/simulation runs necessary to provide meaningful evidence of AV safety performance statistically equivalent to human-driven non-exempt motor vehicles.