Frontier AI systems are crossing dangerous capability thresholds while the institutions designed to govern them remain stuck in consensus-seeking models built for 1945. The Second International AI Safety Report documents the risks. The question is whether any governance architecture can keep pace.