A string of incidents involving AI agents going rogue presses lawmakers to define who is responsible when lines get crossed.