Negotiate a binding international treaty framework for frontier AI safety governance, establishing: mandatory safety testing and evaluation standards for AI systems above defined compute thresholds; international incident reporting and response mechanisms; a multilateral AI Safety Observatory with independent auditing authority; restrictions on autonomous weapons systems; and a global AI Safety Fund supporting safety research in underrepresented countries.

Treaty design would draw on precedents from nuclear safety governance (IAEA) and chemical weapons verification (OPCW).