- Coalition call. The CEOs of Anthropic, Google DeepMind, and OpenAI used a closed-door G7 lunch in Évian on June 17 to urge the United States to lead a multilateral coalition that would set shared standards for testing, evaluating, and governing advanced AI systems.
- Tension. The meeting came days after the US government imposed export controls on Anthropic’s most capable models — Fable 5 and Mythos 5 — for foreign nationals, a move that French President Emmanuel Macron raised directly in the room as a contradiction of the coalition idea.
- Reception. Canada’s Prime Minister Mark Carney signalled openness to Washington leading the effort; other G7 leaders acknowledged the US “certainly could” lead, a formulation short of commitment.
The meeting, held on the sidelines of the G7 summit in Évian-les-Bains, brought approximately twelve technology executives together with heads of state including President Trump, Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnik, and Secretary of State Marco Rubio. Anthropic CEO Dario Amodei, Google DeepMind CEO Demis Hassabis, and OpenAI CEO Sam Altman presented a broadly aligned vision for international AI governance — an unusual degree of inter-company coordination on a policy question among firms that compete aggressively on capability benchmarks.
What the CEOs proposed
Sam Altman offered the most specific institutional blueprint: an “international forum for discussion” with three components — establishing globally accepted standards for AI testing, providing impartial expert analysis of AI capabilities and risks, and serving as “a venue for cooperation among nations.” Amodei focused on the risk of democratic fragmentation, warning that AI tools could be used by “bad actors” and arguing that disagreements among democratic nations over deployment approaches would weaken their collective position relative to non-democratic actors developing the same technologies. He advocated against any approach that would leave allied countries “divided over AI rollout approaches.”
Hassabis of Google DeepMind participated in the session without issuing separate specific proposals. The three CEOs were aligned on the core ask: rather than allowing AI governance to default to existing multilateral bodies where strategic competitors have influence, the US should organise a democratic coalition around common norms and testing standards first.
The Anthropic export control paradox
Macron raised an immediate complication. The US government had recently directed Anthropic to cut off access to its Fable 5 and Mythos 5 models for all foreign nationals — including those in allied nations — citing national security concerns. Macron observed that restricting democratic allies’ access to frontier AI systems runs counter to the spirit of a US-led coalition built on shared access to AI capabilities and governance. His intervention pointed to a structural tension in the US position: the administration wants allied alignment on AI standards while simultaneously restricting the models that make the standards relevant.
According to CNBC, Carney signalled openness to US leadership of the proposed body. Anthropic had separately, in the days before the Évian summit, proposed a global pause on frontier AI development to allow alignment research to catch up with capability advances — a position that found little support among other executives present but underlines how far outside the conventional industry consensus some of Anthropic’s policy instincts now sit. Whether the G7 lunch translates into a formal institutional proposal, or remains a directional statement, will depend on whether the Trump administration decides to pursue the coalition idea in the months ahead.