AI Giants at the G7: Sam Altman, Dario Amodei, and the High-Stakes Lunch That Revealed a New World Order for Artificial Intelligence
AI Giants at the G7: Sam Altman, Dario Amodei, and the High-Stakes Lunch That Revealed a New World Order for Artificial Intelligence
A working lunch in the French Alps brought together the world's most powerful AI executives and heads of state — but it was the explosive backstory of a U.S. export ban on Anthropic's top models that turned a diplomatic gathering into a geopolitical flashpoint.
For the first time ever, the CEOs of the world's leading frontier AI companies sat shoulder-to-shoulder with heads of state not as guests, but as participants in global governance. And if there's one takeaway from this extraordinary gathering, it's this: the AI race has officially become a geopolitical battleground, and the United States is not entirely sure which side it's on.
What Happened at the G7 AI Working Lunch
The formal theme of Wednesday's session was "ensuring a safe, rapid and effective deployment of artificial intelligence." What emerged was far more dramatic.
Altman, the first CEO to speak, called for the creation of a U.S.-led international forum — a body that would establish globally accepted standards for AI testing, provide impartial risk analysis, and serve as a diplomatic venue for cooperation among democratic nations. Canadian Prime Minister Mark Carney reportedly likened the idea to the Financial Stability Board — the international financial watchdog he once chaired — calling the concept a potentially powerful framework for the AI era.
Amodei and Hassabis went further, jointly proposing a formal U.S.-led coalition that would shape international rules for AI, coordinate on structured access to frontier models, restrict AI chip trade with China, and address cybersecurity and bioterrorism risks together.
OpenAI's global affairs chief Chris Lehane, speaking to reporters after the lunch, said the session produced real momentum: there was "a coalescing amongst the countries and the businesses in the room" around the idea of a democratic standards forum, he said. Non-U.S. leaders in the room acknowledged the U.S. "certainly could play the lead role in working to establish" those standards.
It was an ambitious vision — and it was almost immediately undercut by reality.
The Bombshell That Overshadowed Everything: The Anthropic Export Ban
Five days before this historic lunch, the Trump administration dropped a bombshell that rattled Silicon Valley and allied capitals alike.
On June 12, 2026, at 5:21 p.m. ET, the Commerce Department's Bureau of Industry and Security (BIS) issued an export control directive to Anthropic. The order: immediately suspend access to Fable 5 and Mythos 5 — Anthropic's most advanced AI models — for any foreign national, whether located abroad or inside the United States, including foreign national employees of the company itself.
Because Anthropic had no reliable way to verify the nationalities of its hundreds of millions of users on short notice, the company was forced to do the only thing it could: shut both models down entirely, for everyone, worldwide.
It marked the first time the U.S. government has ever invoked the 2018 Export Control Reform Act to regulate an AI model as an emerging technology — a historic and potentially far-reaching legal precedent.
Why did the government act? Commerce Secretary Howard Lutnick cited national security, specifically fears that the models — particularly Mythos, Anthropic's extraordinarily powerful cybersecurity AI — could be exploited by military or intelligence agencies in China, Russia, or other adversarial nations.
The proximate trigger was a reported jailbreak: another company claimed it had found a way to bypass Fable 5's safeguards, potentially unlocking Mythos's full cyber capabilities. Officials were alarmed.
Anthropic strongly disputed the justification. The company argued the reported jailbreak was "narrow" — capable of bypassing safeguards only in one specific instance, not universally. The company also noted that the same technique could be used on other publicly available models, including OpenAI's GPT-5.5, which faced no similar controls.
"We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people," Anthropic wrote in its public statement.
America's Allies Are Furious — and Worried
Here's what made the ban particularly explosive: it didn't just cut off adversaries. It cut off America's closest allies, too.
The UK, France, Canada, Germany, Japan — none of them were spared. Allied governments and businesses that had come to rely on Anthropic's Mythos model for critical cybersecurity work suddenly found themselves locked out overnight. And when British Prime Minister Keir Starmer raised the issue directly with Trump at the G7, seeking a carve-out for British citizens and businesses, his request was formally turned away. A Trump administration official told reporters that granting exemptions to allies would be "completely illogical" given the purpose of the restrictions.
The refusal landed hard in London. UK AI minister Kanishka Narayan called the restrictions evidence that Britain needs greater "sovereign AI capability." Business secretary Peter Kyle pointed to the need to accelerate domestic technology innovation. "This week, the most advanced AI in the world was cut off for everyone in Britain," Narayan said flatly.
The reaction was even sharper in Brussels. European Union officials were particularly stung because the export controls arrived just as Brussels was preparing to join the Pax Silica — Washington's new alliance to secure global AI chip and critical minerals supply chains. Being hit at the exact same moment wasn't lost on anyone.
"The Anthropic kill switch shows that tech sovereignty was never abstract," said Italian MEP Brando Benifei. "The G7 should not lock allies into competing AI dependencies. Europe must cooperate with the US, Canada and democratic partners, but from a position of strength."
The Atlantic Council's Emerson Brooking put it plainly: "Multiple G7 nations have previously alluded to the need for sovereign AI investment, but there was always an assumption that this would take place alongside access to the U.S. tech stack. Now the U.S. has indicated a willingness to cut off the G7 and even treaty allies from certain AI capabilities."
Meanwhile, French President Emmanuel Macron, who personally invited Altman to the summit, called it a "good thing" that U.S. officials recognized frontier models could be dangerous — but criticized the export control move as a "strictly nationalist" reaction.
Dario Amodei's Extraordinary Balancing Act
Of all the figures at the Évian lunch table, none occupied a stranger position than Dario Amodei.
Just five days earlier, his company's flagship products had been pulled from every user on Earth by order of the very government officials now sitting across from him. And yet there he was, calling on the United States to lead a global AI coalition — the same U.S. that had just effectively nationalized control over his technology.
The timing of Amodei's call for American leadership was not lost on observers. It came after a period of severe friction between Anthropic and the Trump administration. Earlier in 2026, Trump ordered all federal agencies to stop using Anthropic's AI products after the company refused to allow its models to be used for autonomous weapons systems or mass domestic surveillance. Defense Secretary Pete Hegseth subsequently designated Anthropic a "Supply-Chain Risk to National Security" — a classification typically reserved for companies linked to foreign adversaries — and Anthropic is now challenging that designation in federal court.
And yet Amodei, speaking at the G7, argued for structured international access to frontier AI models and a chip trade framework that excludes China — effectively endorsing U.S. leadership even while locked in a legal battle with the U.S. government.
It's a testament to how high the stakes have become that Anthropic's CEO found himself, simultaneously, challenging Washington in federal court and asking Washington to lead the world on AI governance.
As of Wednesday evening, negotiations between Anthropic and Commerce Department officials were still underway, with no resolution announced. Fable 5 and Mythos 5 remain offline.
The Deeper Issue: Who Controls the "Kill Switch" on AI?
The Anthropic episode has crystallized something that European policymakers, cybersecurity experts, and tech sovereignty advocates have warned about for years: American AI platforms carry what amounts to a "kill switch."
With a single letter — delivered at 5:21 p.m. on a Friday — Washington was able to instantly disable a product used by hundreds of millions of people worldwide. Not just adversaries. Not just neutral nations. Allied governments. Allied businesses. Allied researchers. Everyone.
"By locking non-Americans out of its most advanced AI models, Trump may have gone a step further: proving that Washington can shut enemies and allies alike out of the technology of its choosing, overnight," noted Euronews in its G7 coverage.
The episode has given fresh urgency to Europe's "AI sovereignty" agenda. France has already deployed Mistral AI's models to all of its civil servants and committed €655 million to domestic AI infrastructure. Germany is accelerating its own sovereign AI initiatives. The EU is reconsidering how deeply it wants to embed American AI into critical infrastructure.
The paradox is stark: the Trump administration's AI Action Plan calls for establishing American AI "as the gold standard for AI worldwide" and ensuring allies "build on American technology." But when the U.S. government demonstrates it can pull that technology from allied hands at a moment's notice, the incentive to do so weakens considerably.
What the Summit Actually Produced — and Didn't
To be clear about what Wednesday's lunch was and wasn't: it was a conversation, not a negotiation, and it produced no binding commitments.
The G7's track record on AI governance is largely one of principles without enforcement. The 2023 Hiroshima AI Process produced voluntary codes of conduct. The 2025 Canadian presidency yielded similar pledges. Wednesday's lunch extended that pattern.
No deal on restoring Anthropic's model access. No formal AI export control framework. No binding sovereignty commitments. No regulatory announcements.
What it did produce was something arguably just as significant: a visible power map of how AI governance actually works in 2026. The CEOs of the companies building the world's most powerful AI sat at the same table as the leaders of the world's most powerful democracies — not as guests or witnesses, but as active participants whose cooperation those leaders genuinely need.
"It just shows that in order to make credible commitments on AI, heads of state now need the cooperation, if not endorsement, of a handful of private sector executives actually building the technology," said Jessica Brandt, senior fellow at the Council on Foreign Relations.
G7 diplomatic sources told Reuters that allies have raised the idea of a "trusted partner" framework with Lutnick — a scheme that would restore AI access for select allied nations or companies — but no agreement has been reached.
What This Means for You
For everyday Americans — and frankly, for anyone in the world who uses AI tools in their work or daily life — the G7 summit and the Anthropic export ban are a reminder that the AI products we've come to rely on are not just consumer software. They are geopolitical assets, subject to the same national security logic as weapons systems, semiconductors, and satellite technology.
The decisions being made right now — about what AI can be exported, who can access it, what standards it must meet, and who gets to set those standards — will shape the global technology landscape for decades to come.
And they're being made, in part, over working lunches in French Alpine towns.
Key Takeaways
- OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, and Google DeepMind CEO Demis Hassabis attended a historic working lunch with G7 leaders, including President Trump, on June 17, 2026.
- Altman proposed a U.S.-led international forum to establish globally accepted AI standards. Amodei and Hassabis called for a formal U.S.-led AI coalition to govern frontier models.
- Just five days earlier, the U.S. Commerce Department ordered Anthropic to disable its two most advanced AI models — Fable 5 and Mythos 5 — for all foreign nationals, in a historic first use of export control law against an AI system.
- The action cut off not just adversaries but U.S. allies, triggering fury in the UK, EU, France, and Canada.
- UK Prime Minister Starmer's request for a British exemption was denied at the summit.
- No binding commitments or policy outcomes emerged from the lunch.
- The episode has supercharged European calls for AI sovereignty and reduced dependence on American technology platforms.
- Anthropic and the Trump administration remain in active negotiations, with Fable 5 and Mythos 5 still offline.
