In this paper, we present a method to assess functional safety of architectures for Automated Driving Systems (ADS). The ISO 26262 standard defines requirements and processes in support of achieving functional safety of passenger vehicles, but does not address in particular autonomous driving functions. Autonomous driving will bring with it a number of fundamental changes affecting functional safety. First, there will no longer be a driver capable of controlling the vehicle in case of a failure of the ADS. Second, the hardware and software architectures will become more complex and flexible than those used for conventional vehicles. We present an automated method to assert functional safety of ADS systems in the spirit of ISO 26262 in light of these changes. The approach is model-based and implemented in the QuantUM analysis tool. We illustrate its use in functional safety analysis using a proposed practical ADS architecture and address, in particular, architectural variant analysis.