Abstract
The complexity of dynamics in AI techniques is already approaching that ofcomplex adaptive systems, thus curtailing the feasibility of formalcontrollability and reachability analysis in the context of AI safety. Itfollows that the envisioned instances of Artificial General Intelligence (AGI)will also suffer from challenges of complexity. To tackle such issues, wepropose the modeling of deleterious behaviors in AI and AGI as psychologicaldisorders, thereby enabling the employment of psychopathological approaches toanalysis and control of misbehaviors. Accordingly, we present a discussion onthe feasibility of the psychopathological approaches to AI safety, and proposegeneral directions for research on modeling, diagnosis, and treatment ofpsychological disorders in AGI.