Macron, Modi warn U.S. could cut off access to AI models overnight

The shock inside the AI boom is no longer about speed or cost. It is about leverage: if the most advanced models sit on American platforms, Washington or a U.S. company can sever access overnight, and allies now want safeguards before that power is used.

Anthropic’s decision on June 12 to disable access to its Fable 5 and Mythos 5 models after a U.S. export-control directive restricted access for foreign nationals turned that fear into a live case study. The move hung over the G7 meetings in Evian-les-Bains, where leaders and technology executives from OpenAI, Anthropic and Google DeepMind weighed how to keep frontier AI available to close partners without giving up national-security controls.

Emmanuel Macron used the summit’s closing stretch to press the issue. On June 17, he said he expected progress in the coming weeks on broadening access to Anthropic’s top models and warned that if U.S. companies cut off access “like a light switch,” their value could decline. The French president also pushed deeper cooperation with India on “cooperative AI,” and France and India adopted an Innovation Roadmap 2030 aimed at expanding collaboration across AI and other emerging technologies.

The immediate policy question is whether the G7 can build a “trusted partners” framework that would let select allies keep access to advanced U.S. models even as export rules tighten. That idea would give governments and companies a way to preserve access to cutting-edge systems from firms such as Anthropic while still reflecting the Biden-era-era? no, current U.S. national-security posture toward foreign use of frontier AI. The debate underscores a bigger shift: advanced AI is being treated less like ordinary software and more like strategic infrastructure, with chips, compute and model access all becoming parts of the same diplomatic bargain.

Canada has already begun moving in that direction. Prime Minister Mark Carney said the Anthropic restrictions showed why countries need to diversify and secure “unhindered access to AI.” Ottawa outlined its National Artificial Intelligence Strategy: AI for All on June 4, framing AI as critical infrastructure and pledging to reduce reliance on U.S. systems. For allies, the lesson from the Anthropic blackout is blunt: sovereignty in the AI era will not be defined only by who builds the best model, but by who controls the switch.