The disclosure provides a method for operating an automatically driving vehicle, wherein application instances are executed over several computational nodes, wherein recognized faults are reacted to by switching to redundant application instances and then reconfiguring the configuration to restore specified redundancy conditions and/or segregation conditions, wherein the vehicle is transitioned to a safe state using at least one failover apparatus when at least one specified redundancy condition and/or at least one segregation condition cannot be met by the reconfiguration, and/or a specified time for reconfiguration is exceeded, and/or an unrecoverable malfunction has been recognized, wherein the at least one failover apparatus plans an emergency trajectory using a trajectory planner, wherein sensor data are detected via separate signal lines and supplied to the at least one failover apparatus, and wherein control signals are generated and transmitted via separate control lines to an actuator system of the vehicle.