Paul Christiano, a board member of the OpenAI Foundation since September 2026, estimates that there is about a 10% chance an AI system will take control within the next 6 to 12 months. Two months earlier, he put the one-year risk at 4% and the three-year risk at 15%, without protective measures. He now estimates that the probability could approach 50% over three years if AI laboratories fail to take preventive action. Christiano founded the Alignment Research Center and helped develop reinforcement learning from human feedback, known as RLHF.
Source: Read the original article


