G7 Confronts America's AI 'Kill Switch': What Happened When World Leaders Sat Down with Altman, Amodei, and Hassabis

The 52nd G7 summit in the French alpine town of Évian-les-Bains was always going to be about artificial intelligence. Host President Emmanuel Macron had placed AI at the top of the agenda, securing the first-ever joint attendance of OpenAI’s Sam Altman, Anthropic’s Dario Amodei, and Google DeepMind’s Demis Hassabis at a G7 leaders’ summit . But a bombshell US decision days before the summit transformed the gathering from a high-level policy discussion into a tense geopolitical negotiation. On June 13, 2026, the Trump administration ordered Anthropic to disable its most advanced models, Fable 5 and Mythos 5, for all foreign nationals, citing national security . The move confirmed Europe's worst fear: that America could, at any moment, flip a "kill switch" on the AI models its allies rely on .

The Anthropic Block: How a 'Kill Switch' Dominated the Summit

The timing could not have been more dramatic. The US government's directive to Anthropic to suspend foreign access to Fable 5 and Mythos 5 landed just two days before the summit began . The export controls represented the most significant step to date to restrict access to cutting-edge AI models . European leaders arrived in Évian with what Euronews described as "a bad aftertaste" and a newly urgent question: if the US could unilaterally sever access to critical AI infrastructure overnight, what did that mean for the economies, security, and sovereignty of its closest allies?

Canadian Prime Minister Mark Carney warned even before the summit that the restrictions exposed the dangers of overreliance on a handful of American AI providers . Inside the summit's working lunch on innovation and AI, the issue dominated the conversation. Macron reportedly warned the gathered leaders and CEOs—including US President Donald Trump—that if the US "from one day to the next can turn off the switch," it would not only harm European customers but damage the American AI companies themselves .

The 'Trusted Partners' Proposal: A Diplomatic Workaround

In response to the crisis, G7 leaders began discussing a "trusted partners" scheme that would carve out exceptions for allied nations—including France, the UK, Germany, and Japan—to continue accessing advanced US AI models even under export restrictions . The proposal, discussed by G7 leaders based on reports from three diplomatic sources, represented an attempt to square the US national security rationale with the economic and strategic needs of its allies .

By the summit's final day, Macron publicly stated he expected "progress in coming weeks on broadening access" to Anthropic’s models . A joint G7 leaders' statement pledged closer coordination on frontier AI risks and opportunities, tasking finance officials, regulators, and cybersecurity experts with assessing how advanced AI models could impact financial stability, productivity, and labor markets .

Europe's Tech Sovereignty Push Gains Urgent Momentum

The Anthropic export block handed a powerful argument to European officials who had already been pushing for greater digital independence. The European Commission had unveiled its European Technological Sovereignty Package on June 3, 2026, a sweeping set of measures designed to reduce the bloc’s "near-total reliance" on American and Asian companies for cloud infrastructure, semiconductors, and AI . The package's four pillars include:

• The Cloud and AI Development Act (CADA), introducing formal sovereignty risk assessments into public procurement for the first time .
• Chips Act 2.0, aimed at deepening Europe's semiconductor ecosystem with an estimated €120 billion in additional investment .
• An EU Open Source Strategy to reinforce Europe’s autonomy across the entire technology stack .
• A Strategic Roadmap for Digitalisation and AI in Energy .

The total investment picture is substantial: the European Commission's communication estimates approximately €200 billion by 2036 to expand data-center capacity, €100 billion for cloud and AI leadership initiatives including AI gigafactories, and €2 billion over seven years for the open-source strategy . European Commission President Ursula von der Leyen, speaking before the summit, emphasized that AI gigafactories would allow European startups to test, train, and improve AI models inside the EU rather than being forced to move elsewhere .

The timing made the sovereignty package the summit's unspoken subtext. As one analysis put it, the formal theme of the AI discussions was "Ensuring a safe, rapid and effective deployment of artificial intelligence," but the actual debate was whether the rest of the world could accept that America's AI companies operate under conditions defined solely by the US executive branch .

A Historic First: Joint G7 Approach to Child Safety Online

Beyond the geopolitics, the summit achieved a concrete win on child protection. G7 Digital Ministers, meeting in Paris on May 29, 2026, had already agreed to a common set of principles to protect children and young people from online harm for the first time, including commitments on age verification, protection of minors from the design stage of digital services, and tackling illegal content .

At the Évian leaders' summit, a dedicated working lunch on child safety was chaired by Mistral AI CEO Arthur Mensch. The top US AI CEOs and smaller AI labs from each G7 nation were also present . G7 leaders issued a joint call for tech companies to develop tools ensuring minors' safety online . The session marked the first time the heads of OpenAI, Anthropic, and Google DeepMind sat at a leaders' summit table to address a specific societal harm linked to their technologies .

What the Evian Summit Signals for the Future

The Évian-les-Bains summit will be remembered as the moment AI governance moved decisively from regulatory working groups to the main stage of geopolitical summitry. The simultaneous presence of Altman, Amodei, and Hassabis alongside heads of state was not just symbolism—it reflected a structural shift in who governs artificial intelligence . The summit extended the Hiroshima AI Process launched in 2023, but the crisis atmosphere created by the Anthropic export controls ensured that future G7 AI discussions will be inseparable from questions of sovereignty, trust, and the concentration of power in a few American labs .

For Europe, the summit validated a policy direction already in motion. As one analysis noted, the kill-switch episode "provided yet another boost" for the tech sovereignty agenda . For the US and its AI industry, the message from allies was clear: they want American AI, but they cannot accept a future where access depends entirely on Washington's discretion .