The G7 AI summit opens today in Évian-les-Bains, a lakeside resort town in the French Alps, and for the first time in the forum's fifty-two-year history, the chief executives of the world's most consequential artificial intelligence companies are sitting at or near the leaders' table. Sam Altman of OpenAI, Dario Amodei of Anthropic, and Demis Hassabis of Google DeepMind have all confirmed attendance, invited by French president Emmanuel Macron as part of his effort to make AI the defining issue of France's G7 presidency. The optics are deliberately historic. But beneath the photogenic lakeside staging, something less ceremonial is taking place. The governments of the United States, the European Union, Canada and their G7 partners are arriving with fundamentally incompatible visions for how the world should govern and benefit from AI, and no amount of diplomatic choreography can paper over the fractures that have emerged in the weeks leading up to this summit.
The Week That Changed the Frame
The volume of national AI declarations that preceded the G7 AI summit was itself a signal. Within four days in early June, three major Western powers published competing strategies. On June 2, the White House published an executive order framing advanced AI as a national security priority, emphasising American dominance and the export of what officials call the "American AI technology stack" to allied and developing nations. The goal, stated plainly, was to ensure that the world ran on US infrastructure rather than Chinese alternatives.
The European Commission followed on June 3 with its European Tech Sovereignty package, proposing hundreds of billions in public investment toward EU-owned chip manufacturing, cloud infrastructure and open-source AI alternatives to US-built models. Ursula von der Leyen put it directly: "We cannot afford to depend on others for the technologies that keep our hospitals running, our energy grids stable and our services secure."
Canada arrived on June 4 with its own national AI strategy, anchored by a Sovereign Tech Alliance it had launched with Germany in February and a broader pitch to what prime minister Mark Carney has called the "Middle Powers" movement. Canada has also backed the acquisition of Germany's Aleph Alpha by its domestic AI champion, Cohere. Three major democracies, three strategies, one week. The G7 AI summit opens into that divergence.
The Rivals in the Room
The attendance of Altman, Amodei and Hassabis adds a dimension no previous G7 has had to contend with. These three men lead organisations preparing two of the largest technology listings in history, and between them they control the most capable AI systems in existence.
Both Anthropic and OpenAI filed S-1 registration statements with the US Securities and Exchange Commission in the first week of June. Anthropic submitted its paperwork on June 1, following a $65 billion funding round that valued the company at $965 billion. OpenAI followed on June 8, with Goldman Sachs and Morgan Stanley as lead underwriters and a valuation reported to exceed $1 trillion at listing.
The timing produces an unusual dynamic at the G7 AI summit. Both companies are simultaneously warning governments about the existential risks of advanced AI while building cases for public-market investors that their technology is the most valuable asset in modern financial history. Appearing at a multilateral summit focused on governance gives both firms a chance to demonstrate seriousness without slowing down.
OpenAI's chief global affairs officer Chris Lehane said before the summit that the company expects a package of voluntary commitments to emerge from Évian, with youth safety and frontier biosecurity risks at the top of Altman's agenda. Anthropic and Google DeepMind confirmed attendance but declined to specify what their executives would raise. The last time all three CEOs shared a stage, at India's AI Impact Summit in February, they managed to make a ceremonial photograph go viral for the wrong reasons. It is unlikely to be forgotten in Évian.
The Pause Nobody Is Taking
Ten days before the G7 AI summit opened, Anthropic published what may be its most consequential policy document to date. Written by cofounder Jack Clark and Anthropic's head of research, Marina Favaro, the paper called for a globally coordinated pause on frontier AI development, arguing that AI systems were accelerating their own improvement at a pace that existing safety frameworks could not match.
The paper disclosed that as of May 2026, more than 80 percent of code merged into Anthropic's own production systems was authored by Claude. The company is not a bystander to the dynamic it is warning about. The proposal was conditional: Anthropic said it would pause only if other frontier labs did the same under verifiable conditions, and it acknowledged that US officials regard any slowdown as a strategic gift to China. The paper's most candid passage admitted that a bad actor could use the auspices of a coordinated slowdown to jump ahead in secret.
No pause is expected to emerge from the G7 AI summit. Washington has been explicit in its opposition to any multilateral commitment that could constrain American competitive advantage. But Anthropic's paper has shifted the frame of the conversation. The executives arriving in Évian are not just pitching their products. They are also, in Anthropic's case, publishing documents that say those products may be moving too fast to govern. As Blumefield has tracked throughout 2026, this tension between capability racing and safety advocacy has become the defining contradiction of the AI industry's public posture.
What the Alps Can Actually Deliver
The G7's track record on AI governance is instructive. The bloc launched the Hiroshima AI Process in 2023 under Japan's presidency, producing international guiding principles and a voluntary code of conduct for AI developers. Canada's 2025 presidency deepened those commitments with pledges on AI adoption in public services. Two years of summitry have yielded principles and pledges but no enforceable rules.
French officials have privately acknowledged that the 2026 edition will likely follow the same pattern. Washington has signalled opposition to any language that could constrain American AI companies, and the US position that AI governance is best handled through voluntary industry commitments rather than binding regulation is not going to shift at a lakeside resort in the Alps.
What Évian can do is surface the fracture. The EU, Canada and their growing constellation of middle-power partners, including South Korea, Brazil and Germany, are building something that looks increasingly like a parallel AI ecosystem, one oriented toward sovereignty and public infrastructure rather than Silicon Valley incumbency. The EU signed an AI cooperation agreement with South Korea on June 10. The European Commission's digital envoy traveled to Brazil the following day to formalise a digital partnership. Canada's Sovereign Tech Alliance with Germany is expanding.
The Stakes Beyond the Summit
Whether that coalition can produce alternatives capable of competing with OpenAI, Anthropic and Google DeepMind is a question that will take years to answer. The numbers alone illustrate the gap. Anthropic's reported valuation of $965 billion and OpenAI's anticipated trillion-dollar listing would together exceed the market capitalisation of most G7 member states' entire technology sectors. Europe's largest homegrown AI challenger, Mistral, is in funding talks at a valuation of around €20 billion.
Macron understands this arithmetic, which is why France has spent its G7 presidency trying to position itself as a bridge rather than a dissident. Inviting Altman to the summit was a deliberate signal that Paris wants a seat at the American AI table even as Brussels builds alternatives. The contradiction is sustainable for now. Whether it remains so after the listings close, the models grow more powerful, and the next round of industrial policy choices must be made is a different question entirely.
What is certain is that the G7 AI summit will be remembered not for the commitments it produced but for the moment it made the fracture visible. The world's most powerful AI companies and the world's most powerful democracies are in the same room, pursuing objectives that no longer fully align. The summit lasts three days. The tension will last considerably longer.