We argue that the critical threats of AI deskilling and addiction, resulting from over-reliance (excessive dependence) and overconfidence (trust) in GenAI assistants, have been overlooked, with the scope of AI safety work being largely limited to certain clusters of risks. 2/5
