At a June 17 G7 working lunch in Évian-les-Bains, France, President Emmanuel Macron put Dario Amodei of Anthropic, Demis Hassabis of Google DeepMind and Sam Altman of OpenAI at the same table as heads of state after a U.S. Commerce Department directive led Anthropic to cut off its Fable 5 and Mythos 5 models worldwide. The episode matters because European companies and public bodies had built services around those models and are now asking whether advanced AI systems can be relied on when access can be changed by a foreign government.

The June 12 directive ordered Anthropic to block its most capable models for any "foreign national," according to the source account. Because nationality checks are difficult at API scale, Anthropic’s practical response was a global shutdown of Fable 5 and Mythos 5 access. European users affected by the move included businesses and public institutions that had built the models into operations.

The AI session at the G7 summit ran against that backdrop. Alongside Amodei, Hassabis and Altman, the guest list included Marc Benioff of Salesforce, Meta AI chief Alexandr Wang and labs from Europe and allied countries, including Mistral, Synthesia, Black Forest Labs, Domyn and Sakana AI. The U.S. delegation included President Donald Trump and Secretaries Scott Bessent, Howard Lutnick and Marco Rubio; European participants included Macron, German Chancellor Friedrich Merz, European Commission President Ursula von der Leyen and U.K. Prime Minister Keir Starmer.

The formal agenda centered on safe, fast and effective AI deployment. The sharper policy question was whether non-U.S. allies can accept a market in which American AI companies remain subject to U.S. executive decisions that can alter access for customers abroad.

Europe Tests AI Dependence

The shutdown turned digital dependency from a strategic worry into an operational risk. If access to frontier models can disappear without warning, AI use in banking, health administration, public services, security, research and commercial software becomes a continuity problem, not only a technology purchase.

For Europe, the answer being tested is not limited to better contracts. Leaders want durable access rights for trusted partners, rules that reduce sudden cutoff risk, a role in deciding where compute, power and chips are located, and domestic capacity through a EUR 420 billion package, AI gigafactories and the CADA effort cited in the summit material. They also pressed for child and youth protections, including age limits and safety by design.

June Directive Recast Summit

The timeline compressed the debate. On June 12, Washington issued the export-control directive. From June 15 to June 17, leaders gathered for the G7 summit in Évian-les-Bains. On June 17, Macron devoted a working lunch to AI, placing company executives beside elected leaders days after the service interruption.

The CEOs arrived with overlapping proposals. Amodei called for a U.S.-led coalition of democratic states with structured access for trusted partners, chip and component trade that excludes China, and cooperation on cyber, bioterrorism and intelligence risks. Hassabis backed a Western coalition and described the moment as one of the most critical in human history. Altman called for an international forum to set testing standards, saying no single lab should make the decisions alone.

“Democracies must not give in to the temptation to splinter.”

— Dario Amodei, Anthropic chief executive

Control Over Access Remains Split

It is not yet clear whether the U.S. ban will be narrowed, whether a trusted-partner exemption can be made to work at API scale, or what binding guarantees companies could give foreign customers while they remain under U.S. law. As of June 23, no reversal had been confirmed.

It is also unclear how quickly European compute projects, gigafactories or self-hosted models could reduce exposure to sudden loss of access. The CEOs can propose access systems and standards, but the Évian dispute centers on a power they do not fully hold: U.S. control over export rules.

September Talks Test Guarantees

Macron’s proposed platform for Western democracies is expected within a month, and leaders are slated to reconvene in September. The next test is whether governments can define trusted-partner access, shared testing standards and child-safety principles in language that U.S. agencies, European institutions and AI companies can all follow.

For companies and public bodies using frontier models, the practical next step is risk planning. Buyers will be watching for contractual guarantees, backup providers, self-hosting options and open-weight systems that can keep services running if another export-control order changes access again.

Key Questions

What happened at the G7 AI lunch?

Macron hosted a June 17 working lunch on AI with G7 leaders and major AI executives, including Amodei, Hassabis and Altman. The meeting followed a U.S. export-control directive that led Anthropic to block Fable 5 and Mythos 5 worldwide.

Why did Anthropic block the models worldwide?

The directive applied to access by any "foreign national." The source account says Anthropic could not reliably verify nationality at API scale, making a global shutdown the practical response.

What does Europe want from the AI chiefs?

European leaders want reliable access, protection from sudden shutoffs, a trusted-partner scheme, a say over compute and chip infrastructure, stronger domestic capacity and child-safety rules.

Can the CEOs guarantee access?

No clear guarantee has been confirmed. The executives can offer access systems and testing forums, but U.S. export policy is controlled by Washington, not by the AI labs alone.

What is the next milestone?

A Western-democracy AI platform is expected within a month, with leaders due to reconvene in September to work on access rights, safety principles and trusted-partner rules.

Source: Thorsten Meyer AI