G7 Leaders Want American AI — But Not the Kill Switch
When the Trump administration issued an unprecedented export control directive on June 12, 2026, ordering Anthropic to suspend all access to its two most advanced AI models for any foreign national anywhere in the world, it did more than disrupt enterprise software contracts. It confirmed a fear that European and allied policymakers had been quietly raising for years: that American AI infrastructure could be switched off overnight, without warning, by a single government order. Five days later, the world's most powerful democratic leaders — and the CEOs of the world's most powerful AI labs — gathered in Évian-les-Bains, France, to figure out what to do about it.
The Anthropic Blackout: What Actually Happened
At 5:21 PM ET on June 12, 2026, the U.S. government issued an export control directive ordering the suspension of all access to Anthropic's Fable 5 and Mythos 5 models by any foreign national — whether located inside or outside the United States, and including foreign national Anthropic employees. The order was sweeping in scope and came with a notable absence of detail. According to Anthropic's official statement, the government's letter "did not provide specific details of its national security concern." Anthropic's own understanding was that the government believed a method to jailbreak Fable 5 had been discovered — a model that had been publicly launched just three days before the directive was issued.
The practical result was blunt: Anthropic had no real-time mechanism to restrict access by nationality. Rather than attempt selective compliance, the company made the decision to take both models offline entirely. As Anthropic stated directly: "The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance." That meant paying enterprise customers, U.S.-based users, and allied governments lost access simultaneously — not because they were the target of the order, but because complying with it in any other way was technically unfeasible on short notice.
Anthropic disputed the government's reasoning. The company said publicly: "We believe this is a misunderstanding and are working to restore access as soon as possible," and stated that the action does not adhere to principles of transparent and fair statutory process. According to multiple analysts tracking the situation, this appears to be the first time the United States government has ever issued an export control directive specifically targeting access to a large language model — marking a significant and largely unprecedented escalation in the federal government's assertion of authority over AI systems.
The Fable 5 incident was not Anthropic's first brush with Washington's national security bureaucracy. In early March 2026, the U.S. Department of Defense had already classified Anthropic as a "supply chain risk" — a designation that prompted the company to file a lawsuit against the Pentagon to block placement on a national security blacklist. In federal court filings submitted at that time, Anthropic's CFO warned that the government's actions could reduce the company's 2026 revenue by multiple billions of dollars. The company's Chief Commercial Officer noted that negotiations with financial institutions worth roughly $180 million combined had already been disrupted as a result of the Pentagon situation. Between December 2025 and January 2026 alone, Anthropic had seen a fourfold increase in annual recurring revenue from public sector customers, with projected public sector revenue for 2026 exceeding half a billion dollars — a commercial trajectory now directly exposed to geopolitical risk.
G7 in Évian: AI Sovereignty Moves to the Main Stage
The 52nd G7 Summit, held June 15–17, 2026, in Évian-les-Bains and hosted by French President Emmanuel Macron, was already slated to address AI governance on its final day. The Anthropic blackout ensured the conversation arrived with considerably more urgency than anyone had planned for.
France had placed AI at the heart of its G7 presidency, with the working lunch on June 17 formally themed "Ensuring a safe, rapid and effective deployment of artificial intelligence." What made the session historically significant was who showed up. According to CNBC, the working lunch was joined by the CEOs of Anthropic, OpenAI, and 11 other technology companies. For the first time, the heads of the three most consequential AI laboratories — OpenAI CEO Sam Altman, Google DeepMind CEO Demis Hassabis, and Anthropic CEO Dario Amodei — appeared simultaneously before heads of state at a G7 summit. The symbolism was hard to miss: frontier AI had become a matter of state.
Macron met directly with both Amodei and Altman ahead of the working lunch, according to Bloomberg, with the explicit goal of pushing for European access to leading-edge AI models. Meanwhile, U.S. Commerce Secretary Howard Lutnick pitched a potential path forward: a "trusted partners" framework, under which vetted allied nations and approved companies would be granted exemptions from the kind of sweeping restrictions imposed on June 12. G7 leaders discussed the proposed framework on June 16, with the concept representing an attempt to square the circle between U.S. national security concerns and allied nations' demand for stable, uninterrupted access to American AI technology.
India, invited as a partner nation to the summit alongside Brazil, Kenya, and South Korea, was also part of the broader discussions — reflecting the extent to which concern over AI access had spread well beyond Europe into the Global South.
Europe's Kill-Switch Fear Becomes a Documented Reality
For European policymakers, the June 12 directive was less a surprise than a confirmation. The fear that U.S. AI companies could function as a de facto kill switch over critical digital infrastructure had been circulating in EU policy circles for some time. What changed on June 12 was that it stopped being hypothetical.
The timing compounded the political difficulty. According to Euronews, the U.S. export restrictions landed precisely as Brussels was preparing to join the Pax Silica — Washington's alliance to secure global supply chains for AI chips and critical minerals — making the blackout an awkward moment in what was supposed to be a deepening transatlantic technology relationship.
The European Commission's response was pointed. Thomas Regnier, the Commission's spokesperson for tech sovereignty, addressed the situation directly: "We are a trusted partner. I would challenge you to find a more trusted partner than Europe."
The EU's concern extended beyond general access. According to reporting from Saasultra, the European Union was specifically seeking access to Anthropic's Mythos model — an AI system designed to detect vulnerabilities in computer code — to assess its implications for European cybersecurity infrastructure. The suspension of that model raised direct questions about whether European governments could rely on American AI tools for sensitive infrastructure assessment work, or whether doing so exposed them to precisely the kind of unilateral disruption that had just occurred.
Cohere CEO Aidan Gomez articulated the broader ambition that the episode accelerated. Speaking at the summit, Gomez called for efforts "to expand our sovereign AI ecosystem partnerships beyond Canada and Germany to include all G7 nations — and companies — establishing a global standard that guarantees ownership of models, data, and local compute." The statement captured a direction of travel that the Anthropic blackout had done much to accelerate: a push toward AI arrangements that preserve allied access not through American goodwill, but through structural guarantees embedded in international agreements and sovereign infrastructure.
Why This Matters Beyond Geopolitics
It would be easy to read the Évian summit and the Anthropic blackout as purely a story about great power competition and diplomatic maneuvering. But the underlying dynamics have direct implications for any organization — government, enterprise, or individual — that has built workflows, products, or decision-making processes on top of American AI platforms.
The June 12 event demonstrated that even paying enterprise customers in allied countries are not insulated from sudden, policy-driven access interruptions. The disruption was not caused by a technical failure, a cyberattack, or a commercial dispute — it was the downstream consequence of a national security directive issued with no advance notice and no stated specific justification. For organizations that depend on frontier AI models for productivity, research, or operational continuity, that is a material risk that the Évian discussions have now elevated to the level of formal international negotiation.
The trusted partners framework being discussed at the G7, if it advances, could eventually provide structural protections for allied-nation access. But as of the summit's close on June 17, no formal agreement had been announced, and the gap between the framework's ambition and its current status as a diplomatic concept remained considerable. Meanwhile, the broader pattern — of U.S. authorities asserting export control authority over AI model access — shows no sign of reversing.
For more tech news, visit our news section.