TL;DR
At the June 17 G7 AI lunch in Évian-les-Bains, European leaders met Dario Amodei, Demis Hassabis and Sam Altman days after a U.S. export-control order forced Anthropic to disable Fable 5 and Mythos 5 for users worldwide. The confirmed fallout is a new European push for durable model access, trusted-partner rules and more sovereign compute; it remains unclear whether CEOs can provide guarantees while Washington controls export decisions.
European leaders used a June 17 working lunch at the G7 summit in Évian-les-Bains to address a new AI dependency risk: a U.S. export-control order had just led Anthropic to disable Fable 5 and Mythos 5 worldwide, putting Dario Amodei, Demis Hassabis and Sam Altman at the same table with heads of state as Europe sought firmer access to frontier models.
In a June 12 statement, Anthropic said the U.S. government, citing national security authorities, directed it to suspend access to Fable 5 and Mythos 5 by any "foreign national," including people inside the United States and Anthropic’s own foreign-national employees. The company said filtering users by nationality in real time was not workable, so it removed access for all customers while other Anthropic models stayed available.
The June 17 lunch brought Amodei together with Google DeepMind CEO Demis Hassabis and OpenAI CEO Sam Altman, plus other tech leaders, including representatives from Mistral, Salesforce, Meta AI, Synthesia, Black Forest Labs, Domyn and Sakana AI, according to the source material and Business Insider/Politico reporting. The government side included French President Emmanuel Macron, German Chancellor Friedrich Merz, European Commission President Ursula von der Leyen, U.K. Prime Minister Keir Starmer and U.S. President Donald Trump.
The executives offered broad alignment rather than a reversal of the order: Amodei pressed for a U.S.-led coalition of democratic states, Hassabis backed Western coordination, and Altman called for an international forum to set testing standards. European leaders were focused on access rights, trusted-partner rules, limits on sudden shutdowns, local compute, chips, power and stronger child-safety principles.
Évian and the fallout: what Europe actually wants
For the first time, Amodei, Hassabis, and Altman sat with heads of state — five days after Washington switched Anthropic’s models off worldwide. Europe’s question: can you rely on models a foreign cabinet can shut down by decree?
The dilemma: what Europe wants from the three CEOs, the three can’t deliver — because they don’t hold the switch, Washington does. Macron’s platform is the right answer, but no fix for a decade-old infrastructure gap. The only answer that doesn’t depend on someone else’s goodwill: your own models, your own compute, open weights you can self-host.
Europe Weighs Model Access Risk
The shutdown turned Europe’s AI dependency debate from a policy concern into an operational risk. If a bank, hospital, ministry or software company builds workflows around a frontier model, a foreign export-control decision can interrupt service, incident response, code review, research or customer support with little warning.
That is why the European demand is broader than model access. Leaders want a say in where data centers are built, how power and chips are allocated, and which allies qualify for protected access. They also want fallback capacity through European models, open weights that can be hosted locally and compute infrastructure not fully exposed to a single foreign cabinet decision.
AI model hosting server
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Shutdown Reframed The G7 Lunch
The Évian lunch followed years of European concern about dependence on U.S. cloud platforms, chips and foundation models, but the Anthropic order gave the issue a specific test case. The source material says Macron linked the response to a €420 billion technology-sovereignty package, AI gigafactories and CADA, with child and youth safety also on the summit list.
Amodei’s position was awkward: he asked democratic countries not to split into rival AI blocs only days after his own company was forced to remove its most capable models from European users. Altman’s message pushed authority away from labs, while Hassabis framed the moment as a major turning point for AI governance. None of those positions, on their own, answered Europe’s demand for guaranteed access.
“resist the temptation to splinter”
— Dario Amodei, Anthropic CEO
enterprise AI compute hardware
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Guarantees Remain Outside CEO Control
The key unknown is whether Washington will change the directive, issue country carve-outs or create a trusted-partner license that gives European and allied users stable access. As of June 23, no reversal had been announced in the source material.
The technical basis for the order also remains partly opaque. Anthropic says the government’s concern appears to involve a narrow jailbreak and says other public models can produce similar vulnerability-finding outputs. U.S. officials have cited national security concerns, but the full evidence has not been released publicly.
It is also unresolved whether contracts with AI labs can protect customers from future export-control actions. The central tension is that Amodei, Hassabis and Altman can shape access policies inside their companies, but they do not control U.S. national security law.
secure AI data storage solutions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
September Talks Test Access Rules
The next step identified in the summit fallout is a Macron-led platform due within a month, followed by another meeting of Western democratic leaders in September. Those talks are expected to test whether trusted-partner access, cyber-defense cooperation and child-safety rules can be written into a usable framework.
Anthropic is expected to keep seeking a path to restore Fable 5 and Mythos 5 access. European governments, meanwhile, will use the shutdown as evidence for faster investment in domestic models, data centers and open systems that can keep running if foreign access rules change again.
AI chipsets for data centers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What happened at the Évian G7 AI lunch?
European leaders and top AI executives met on June 17 during the G7 summit to discuss AI deployment, growth, resilience and safety. The recent U.S. order against Anthropic made access to American frontier models the main political issue around the table.
Why did Anthropic disable Fable 5 and Mythos 5 worldwide?
Anthropic said the U.S. government ordered it to suspend access by any foreign national. Because it could not reliably filter that at API scale, it disabled the models for all customers while leaving other Anthropic models online.
What does Europe want from Amodei, Hassabis and Altman?
European leaders want durable access, guarantees against sudden model shutdowns, a trusted-partner scheme for allies, more control over compute and chips, stronger local AI capacity and child-safety rules built into deployment plans.
Can the CEOs grant those guarantees?
Only partly. They can offer company policies, technical standards and partnership structures, but the Anthropic case shows that U.S. export controls can override commercial access decisions when national security authorities intervene.
What happens after the summit?
The source material points to a platform within one month and a September follow-up among Western democratic leaders. The unresolved question is whether that process produces binding access rules or mostly political language.
Source: Thorsten Meyer AI
