European leaders pressed the heads of Anthropic, Google DeepMind and OpenAI at a June 17 G7 working lunch in Évian-les-Bains for more reliable access to frontier AI systems after a U.S. export-control directive forced Anthropic to shut down its most capable models worldwide, exposing a direct operational risk for European companies and public institutions.

The meeting put Dario Amodei, Demis Hassabis and Sam Altman at the same table as heads of state during a summit session hosted by French President Emmanuel Macron and devoted to artificial intelligence. The talks followed a June 12 U.S. Commerce Department directive ordering Anthropic to block Fable 5 and Mythos 5 for any “foreign national.” According to the source account, Anthropic could not reliably check nationality at API scale and responded with a global cutoff.

The result was immediate: European businesses and public bodies that had built workflows around those models lost access without advance warning or a phase-in period. The episode shifted Europe’s AI debate from long-range industrial policy to a near-term question of service continuity.

At the table, the U.S. side included President Donald Trump and senior cabinet officials. European participants included Macron, German Chancellor Friedrich Merz, European Commission President Ursula von der Leyen and U.K. Prime Minister Keir Starmer. Other invited technology leaders included Marc Benioff of Salesforce, Meta AI chief Alexandr Wang, and representatives from Mistral, Synthesia, Black Forest Labs, Domyn and Sakana AI.

Model Access Becomes Leverage

The shutdown matters because it showed that access to leading AI models can be altered by a foreign government decision, even when customers are outside the country issuing the order. For European users, the risk is no longer theoretical: procurement, public services, financial systems and research teams may depend on models they cannot control.

Europe’s demands now extend beyond better contracts. Leaders want durable access rights for trusted non-U.S. partners, protection from another abrupt cutoff, a role in decisions about compute and chip infrastructure, and stronger rules for child and youth safety. The harder issue is that the chief executives can offer partnerships and standards, but they do not control U.S. export policy.

Models, Methods and Tools for Product Service Design: The Manutelligence Project (SpringerBriefs in Applied Sciences and Technology)

View Latest Price

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

The June 12 Cutoff

The source material says Macron’s AI session came five days after Washington used export controls in a way that directly affected global access to Anthropic’s most advanced systems. The G7’s public framing focused on safe, rapid and effective AI deployment, but the operational question was sharper: whether allies can rely on models governed by U.S. executive decisions.

Amodei and Hassabis argued for a U.S.-led coalition of democratic states, with structured access for trusted partners and chip trade rules that exclude China. Altman backed an international forum for model testing standards, arguing that the future of AI should be shaped by democratic institutions rather than individual labs alone.

“temptation to splinter”

— Dario Amodei, Anthropic chief executive

Guarantees Still Lack Detail

It is not yet clear whether the United States will create a legally durable trusted-partner scheme, whether the Anthropic restrictions will be narrowed, or what guarantees any AI company could give customers while export controls remain outside corporate control. The source account says the ban remained in place after the summit.

Details are also still emerging on how Europe’s proposed sovereignty measures, including a reported €420 billion package, gigafactories and CADA, would translate into near-term model access, compute capacity and self-hosted alternatives.

September Will Test Commitments

The next milestones are Macron’s planned AI platform within one month and a September meeting of Western democratic leaders. Those talks are expected to address trusted-partner access, cyber-defense against China-related risks, common safety principles for children and the future status of the Anthropic restrictions.

For Europe, the main test is whether summit language turns into enforceable access rules, domestic compute capacity and models that European institutions can run without relying on another government’s permission.

Key Questions

What happened at the Évian G7 meeting?

European leaders met with top AI executives, including Dario Amodei, Demis Hassabis and Sam Altman, and pressed them on reliable access to frontier AI models after a U.S. directive triggered a global Anthropic cutoff.

Why did Anthropic shut down Fable 5 and Mythos 5?

The source account says a U.S. Commerce Department directive required Anthropic to block the models for any “foreign national.” Because nationality could not be reliably checked in real time at API scale, Anthropic shut access worldwide.

What does Europe want from the AI companies?

European leaders want durable access, safeguards against another abrupt cutoff, trusted-partner rights for allied countries, more say over compute and chip infrastructure, stronger domestic AI capacity, and child safety rules built into deployment.

Can the AI chiefs guarantee access on their own?

Only partly. They can support standards, partnerships and access programs, but the cutoff showed that final authority over export controls sits with Washington, not the companies.

Has the Anthropic ban been reversed?

No reversal is confirmed in the source material. The next test is whether talks after Évian produce a narrower rule, an allied access scheme or a longer-running split between U.S. policy and European operational needs.

Source: Thorsten Meyer AI