Gewirtz on AI risk as seen from Beijing
Julian Gewirtz on the AI risks that actually scare Beijing
AI summary
Former Senior Director for China and Taiwan Affairs at the U.S. National Security Council Julian Gewirtz said on the 80,000 Hours podcast that the CCP values regime security most, and that "extreme loss-of-control risk" refers both to AI loss of control and to loss of political control. He said Anthropic's Claude Mythos is the most significant event in shaping Beijing's perception of AI risk over the past year or so. He believes a U.S.-China agreement should start from common interests such as neither side wanting AI to lose control, and advance step by step.