Autonomous driving requires adaptivity “on steroids” to serve different mission needs in varying and open driving environments using limited on-board resources, possibly under failures. After a brief overview of the functions that make up an automated driving system, I will describe three major control loops in charge of the adaptivity: basic control, supervisory control, and fleet learning. Basic control needs to react to immediate changes of the road environment, such as lighting and weather, to adjust perception and motion planning and control functions for the given driving conditions. Supervisory control monitors the health of the system and its functional limits. Fleet learning allows vehicles to share new experiences from the field with each other. I will describe current designs and mechanisms used in such systems, pointing out limitations, challenges, and future directions.