The G7 Searches for a Way Around America's AI Wall

The leaders of the world's wealthiest democracies arrived in Evian-les-Bains this week expecting to talk about trade and critical minerals. Instead, much of the summit turned on a single question: who gets to use America's most powerful artificial intelligence.

The shift followed an abrupt move by Washington. On 12 June, US authorities issued an export control directive that suspended foreign access to the most advanced frontier models, citing national security. Anthropic complied within hours, disabling its newly released Claude Fable 5 and Mythos 5 for every user outside the United States, including its own foreign-national staff. We covered that order in Washington Orders Anthropic to Pull Fable 5 and Mythos 5 and the company's scramble to negotiate a way back.

By the time the G7 convened, the restrictions had cut off allies and adversaries alike. That left France, Canada, Japan and the rest of the bloc facing the same wall, and looking for a door.

The trusted partners idea

US Commerce Secretary Howard Lutnick used the summit to pitch a way through. According to reporting from Crypto Briefing, his proposed "trusted partners" framework would create a vetted list of allied nations and approved firms that could regain access to top-tier models from companies such as Anthropic, OpenAI and Google. Countries on the list would be treated as safe destinations. Everyone else stays outside.

The pitch lands awkwardly. A vetted list turns access to AI into a diplomatic lever, and "trusted partner" status into something a government can grant or withhold. Several European officials have privately worried that the arrangement hands Washington new control over technology their own economies increasingly depend on, a tension CNBC described as a fresh front of AI protectionism.

The chief executives in the room

France invited the heads of more than a dozen AI companies to Evian, among them Anthropic's Dario Amodei and OpenAI's Sam Altman. Their presence showed how far the industry has moved into the centre of foreign policy. Canadian Prime Minister Mark Carney pressed for AI safety standards that protect children's data, and said he saw strong interest among like-minded countries in sharing the infrastructure to enforce them. The European Union, which already runs the most detailed AI rulebook of any major economy through its AI Act, arrived wanting cooperation on governance rather than dependence on a single supplier.

No deal yet

The summit closed on 17 June with a communique still being drafted and no binding agreement on access. What emerged instead was a clearer picture of the stakes. Frontier AI has become infrastructure that governments treat the way they once treated oil or advanced weaponry, and the United States now holds the tap. Whether the trusted partners idea becomes a real framework or just leverage will decide how much of the world can build on American models, and on whose terms.