TL;DR
Amodei and Hassabis proposed a US-led AI coalition at the G7. Altman called for an international testing forum. No binding commitments emerged.
At a closed-door lunch in Évian-les-Bains with Trump, Macron, and about a dozen tech CEOs, the rivals who spent the past month fighting their own governments pitched Washington as the leader of international AI governance
Amodei and Hassabis proposed a US-led AI coalition at the G7. Altman called for an international testing forum. No binding commitments emerged.
Anthropic CEO Dario Amodei and Google DeepMind’s Demis Hassabis used a closed-door lunch at the G7 summit on Wednesday to call for a US-led coalition that would shape international rules and standards for artificial intelligence, according to two people with knowledge of the discussions who spoke to CNBC on condition of anonymity. Canadian Prime Minister Mark Carney agreed that the US could lead such a coalition, according to one of those people and a third person familiar with the talks.
The meeting took place in Évian-les-Bains, France, on the final day of the three-day summit. About a dozen tech executives attended the lunch alongside G7 heads of state, including President Donald Trump, who was joined by Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and Secretary of State Marco Rubio.
Amodei told the group that areas of international cooperation should include structured access to frontier AI models and trade in chips and critical components that excludes China, according to one of the sources. He also said countries should cooperate to address the risk of AI in cyber operations, bioterrorism, and intelligence, the source added.
OpenAI CEO Sam Altman, who also attended, called for “an international forum for discussion that establishes globally accepted standards for testing, provides expert and impartial analysis of capabilities and risks, and serves as a venue for cooperation among nations,” according to an OpenAI briefing. Chris Lehane, OpenAI’s global affairs chief, said non-US leaders in the room acknowledged that the US “certainly could play the lead role in working to establish” AI standards.
Anthropic declined to comment on the meeting. Google DeepMind and the Office of the Canadian Prime Minister did not immediately respond to requests for comment, according to CNBC.
The gathering produced no binding commitments or regulatory announcements. Multiple accounts described it as a conversation rather than a negotiation, and the G7’s track record on AI governance, from the Hiroshima AI Process in 2023 to Canada’s 2025 presidency pledges, has so far yielded principles and codes of conduct but no enforceable regulation.
The coalition pitch is notable given the backdrop. Anthropic is locked in negotiations with the Trump administration after the US government imposed export controls on its Fable 5 and Mythos 5 models on 12 June, citing national security concerns over a reported jailbreak. The company disabled both models worldwide to comply, and talks with Commerce Department officials were still underway as of Monday.
Amodei’s call for the US to lead international AI governance came just five days after that same government forced his company’s flagship products offline. The tension underscores how AI labs are simultaneously positioning themselves as indispensable partners to Washington on global technology competition while pushing back against the controls Washington imposes on their own products.
The full list of tech attendees extended well beyond the three major US labs. Mistral’s Arthur Mensch, Cohere’s Aidan Gomez, Black Forest Labs’ Robin Rombach, Synthesia’s Victor Riparbelli, Salesforce’s Marc Benioff, and Meta’s Alex Wang all attended, along with executives from Italy’s Domyn, India’s Sarvam AI, and Japan’s Sakana AI, according to CNBC. The geographic spread reflected France’s effort, as holder of the rotating G7 presidency, to cast the AI discussion as a global rather than purely American conversation.
The cyber dimension loomed large over the proceedings. OpenAI rolled out GPT-5.5 Cyber last month in a limited preview to vetted cybersecurity teams, and Anthropic’s own Mythos model had been restricted to cybersecurity defenders before the export controls pulled it entirely. Both companies have argued that frontier AI models are more useful in the hands of defenders than attackers, a position that the US government’s actions have complicated.
Whether a US-led coalition materialises depends on factors well beyond what happened at a single lunch. The Trump administration has shown it is willing to act unilaterally against AI companies it considers a security risk, which could undermine the collaborative framework Amodei and Hassabis described. Several G7 members, including France and the EU more broadly, have pursued their own AI regulatory paths that do not presuppose American leadership.
What the meeting made clear is that the CEOs of the world’s most powerful AI companies now see international governance as something they need to shape rather than resist. The question is whether the governments sitting across the table, several of which have already imposed their own controls, will let them.
Get the most important tech news in your inbox each week.