In this chapter, we discuss teleoperation systems for connected and automated vehicles and overview the research around making these systems safe and human-centric. In recent years, teleoperation by human operators has become an important modality of supervision for connected and automated vehicles. While automated driving systems continue to mature and handle a rich variety of scenarios, we are continually reminded that unplanned exceptions and deviations from the driving systems' operational design domain – be it unpredictable human behavior or unmodeled kangaroo mechanics – can create potentially unsafe situations for automated vehicles. From decades of research in human factors in air traffic control, we know that we can robustify against automation failure by closely integrating human operators into our automation systems. To this end, teleoperation systems have become of great interest to the connected and automated vehicle research community. By establishing teleoperation systems, we can ensure that when a vehicle experiences automation failure, we have a fallback system where a human operator is able to handle the failure. Moreover, due to the recent introduction of ultra-reliable and low latency communication in 5G networks, we know that wireless communication channels are reliable enough to safely deploy such fallback systems. Since the human operator is remote, we can potentially set up “control towers” for vehicles where a few operators can manage many vehicles. However, to reach this potential, there are several safety and cyber–physical–human system design challenges that need to be addressed. In this chapter, we will introduce these challenges and discuss the recent research that has been conducted to solve these challenges. Furthermore, we will discuss the future research directions that will further enhance the safety of teleoperation systems.