TL;DR
At the June 17 G7 working lunch in Évian-les-Bains, European leaders faced a practical risk: access to U.S. frontier AI models can be cut by a U.S. order. The meeting followed a June 12 directive that led Anthropic to disable Fable 5 and Mythos 5 worldwide.
At a June 17 G7 working lunch in Évian-les-Bains devoted to artificial intelligence, Dario Amodei, Demis Hassabis and Sam Altman sat with heads of state days after a U.S. order forced Anthropic to disable Fable 5 and Mythos 5 worldwide, putting Europe’s reliance on U.S. AI models at the center of the summit fallout.
The June 12 U.S. Commerce Department directive barred Anthropic’s most capable models from access by any "foreign national," according to the source material. Because Anthropic could not verify nationality reliably at API scale, it disabled the models for all customers, leaving European businesses and public institutions that used them without warning or a grace period.
French President Emmanuel Macron devoted the lunch to AI under the official theme of safe, rapid and effective deployment. Around a dozen technology leaders attended, including Marc Benioff of Salesforce, Meta AI chief Alexandr Wang, and representatives of Mistral, Synthesia, Black Forest Labs, Domyn and Sakana AI.
The executives offered coordination rather than unilateral company control. Amodei and Hassabis backed a U.S.-led coalition of democratic states, structured access for trusted partners and chip controls aimed at China. Altman proposed an international forum for model testing standards and said decisions about the technology should sit with democratic institutions, not labs alone.
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 Tests AI Dependence
The shutdown changed the European debate from strategy to operating risk. A public agency, hospital, bank, manufacturer or software team that builds on a frontier model can lose service if access depends on a legal order outside Europe. That risk is separate from model quality, pricing or normal reliability.
European leaders are seeking reliable access, protection against another shutdown, trusted-partner rights for non-U.S. allies, influence over where compute and chips are placed, and child and youth safety rules. The source material also cites a €420 billion European sovereignty package and gigafactory plans. Those aims depend on governments as much as the companies at the table.
Plaud NotePin Voice Recorder, AI Voice Recorder, AI Note Taker, AI Transcribe & Summarize, Support 112 Languages, for Meetings, Lectures, Professionals, Teams, Cosmic Gray, Non-S Version
Plaud Intelligence: Capture conversations in 112 languages and generate accurate transcripts with the Plaud App and Web. Plaud…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Shutdown Shaped The Lunch
The Évian summit ran from June 15 to 17, 2026. The AI lunch signaled a shift in diplomatic practice: companies building frontier systems were seated with national leaders because their products now support economic activity, security work and public administration.
The immediate trigger was the Anthropic order. The source material says Fable 5 and Mythos 5 were taken offline worldwide because the U.S. rule applied to foreign nationals and could not be applied reliably user by user. Anthropic complied but disputed the need for a global outage.
“resist the temptation to splinter”
— Dario Amodei, according to accounts of the lunch
Access Control Systems: Security, Identity Management and Trust Models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Access Guarantees Stay Unresolved
It is not yet clear whether Washington will narrow, lift or replace the Anthropic restriction with licenses for trusted partners. It is also not clear what enforceable assurances European customers could get if U.S. law conflicts with commercial contracts.
The source material does not confirm a binding G7 deal on model access, testing standards, child safety principles or infrastructure placement. The CEO offers remain proposals, and the European demands remain policy aims.
Hermes Agent: The Production-Grade AI Engine: Cut API Costs, Add Model Fallbacks, and Secure Your LLM Layer with LiteLLM
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
September Talks Set The Test
Leaders are expected to meet again in September, with France seeking a platform for Western democracies covering trusted access, cyber-defense cooperation and child safety principles. Before then, the immediate marker is whether the U.S. changes the Anthropic order or leaves the ban in place.
Europe’s longer course is domestic capacity: local compute, AI gigafactories, domestic labs and models that firms or agencies can host themselves. That buildout will take years, so access to U.S. frontier models remains a live policy fight.
AI TOOLS FOR PROFESSIONALS
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What happened at the G7 lunch in Évian?
AI leaders Dario Amodei, Demis Hassabis and Sam Altman joined heads of state for a June 17 working lunch focused on AI. The meeting took place days after a U.S. directive led Anthropic to disable Fable 5 and Mythos 5 worldwide.
Why did Anthropic disable Fable 5 and Mythos 5?
According to the source material, the U.S. directive barred access by foreign nationals. Anthropic said it could not reliably check nationality across API users in real time, so it disabled the models globally.
What does Europe want from the AI CEOs?
European leaders want reliable access, protection from another shutdown, trusted-partner rules, a say over compute and chip infrastructure, stronger child safety measures and more domestic AI capacity.
Can the CEOs give Europe binding access guarantees?
Only partly. The companies can offer contracts, access programs and technical cooperation, but they cannot override U.S. export-control decisions. That is the central tension exposed by the Anthropic shutdown.
What happens after the summit?
Leaders are expected to revisit the issue in September. The near-term question is whether the Anthropic ban changes; the longer issue is whether Europe can build enough of its own AI infrastructure to reduce reliance on U.S. providers.
Source: Thorsten Meyer AI
