RaD-AI agents are artificial agents (virtual or robots) that reason intelligently about why, when, and how to rebel and disobey their given commands. The need for agents to disobey contrasts with most existing research on collaborative robots and agents, where the definition of a “good” agent is one that complies with the commands it is given, and that works in a predictable manner under the consent of the human it serves. However, as exemplified in Issac Asimov’s Second Law of Robotics, this compliance is not always desired, such as when it might interfere with a human’s safety. While there has not been much prior research on RaD-AI, we identify main related topics, each of which is studied by a thriving subcommunity of AI, namely: Intelligent Social Agents, Human-Agent/Robot Interaction, and Societal Impacts. In each of these areas, there are research questions relevant to RaD-AI.