OpenAI last week removed Aleksander Madry, one of OpenAI’s top safety executives, from his role and reassigned him to a job focused on AI reasoning.