WORLD LEADERS WANT AMERICAN AI BUT NOBODY WANTS WASHINGTON TO HOLD THE OFF SWITCH The G7 summit in France exposed a fault line running through every major alliance America has. Heads of state from Europe, Canada, Japan and beyond showed up wanting access to frontier AI from OpenAI and Anthropic. They also showed up demanding ironclad guarantees that Washington could never pull the plug. The immediate trigger was the Commerce Department export control directive that forced Anthropic to disable Fable 5 globally, not just for foreign nationals. In a matter of hours, allies who had integrated the model into government workflows discovered what they had long feared: American technology comes with American kill switches. EU leaders have been warning about this for years. Now they have proof. The episode reignited debates inside Brussels about technological sovereignty, accelerated European investment in domestic AI alternatives, and put France’s Macron on the front foot pushing a French and Indian joint AI initiative explicitly framed around trust. The Americans in the room offered a trusted partners carve-out scheme as a compromise. Nobody signed anything. World leaders want American AI. They also want to be sure that if relations ever sour, the servers keep running. Right now they cannot have both. The question of who controls the infrastructure is no longer theoretical. It landed on every G7 leader’s desk this week. Keywords: AI sovereignty, G7 AI summit, US AI export controls, kill switch AI