For decades, the automotive industry has used the notion of controllability as a way of assessing the likelihood that, when subject to a hazardous situation, the vehicle driver avoids an accident. The introduction of semi-and fully-autonomous vehicle systems has the potential to fundamentally alter the driver's tasks and role during all or part of the journey, and challenges the notion of controllability as used in existing automotive risk models. This paper explores the notion of controllability and highlights areas where increasing autonomy may impact the use of controllability in risk assessment.