Could AI enable a small group to gain power over a large country, and lock in their power permanently? Often, people worried about catastrophic risks from AI have been concerned with misalignment risks. In this episode, Tom Davidson talks about a risk that could be comparably important: that of AI-enabled coups.