Western allies arrived at the G7 AI summit unsettled after the Trump administration cut off foreign access to Anthropic's most powerful models, confirming Europe's kill-switch fears and providing yet another boost for its tech sovereignty agenda.
Artificial intelligence is fast becoming a strategic asset, promising economic prosperity and military dominance to those who master it. But as G7 leaders gather in France for the third day of their summit to discuss AI, fears of a potential American “kill switch” have rarely felt more real.
ADVERTISEMENT
ADVERTISEMENT
France placed AI at the heart of its G7 presidency from the outset, convening Western leaders, partner countries and top tech executives to discuss how to ensure "a safe, rapid and effective deployment of artificial intelligence."
French President Emmanuel Macron, facing the G7 spotlight for the last time before his mandate expires next year, has staked part of his legacy on making France an AI powerhouse, positioning Paris as a research and investment hub.
Yet leaders are arriving at the summit's working lunch with a bad aftertaste, following Washington's decision to cut off non-Americans from Anthropic's most powerful models, including the company's own foreign employees.
Pulling the plug
On Friday, Washington imposed export control measures blocking all access to Fable 5 and Mythos 5 by any foreign national. Anthropic consequently had to "abruptly disable" access to both models to comply.
The restrictions, and the speed with which the company applied them, confirmed one of Europe's worst-kept fears: that the Americans could shut down access to their most powerful technology at a moment's notice.
Washington cited national security as its justification. Fable 5 and Mythos 5 are said to have unprecedented capabilities to identify and exploit cyber vulnerabilities, making them a formidable hacking tool.
But being abruptly cut off from the world's most advanced AI models left European and other Western allies flat-footed, delivering the ultimate reality check: access to American technology in the Trump era is no longer a given.
Feeling left out
Brussels argues that, while measures to address the serious risks posed by increasingly powerful AI are justified, they should not discriminate against trusted partners.
"We are a trusted partner. I would challenge you to find a more trusted partner than Europe," said Thomas Regnier, the European Commission's spokesperson for tech sovereignty.
The mood in EU policy circles is particularly sour. The US 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, hitting Europeans at the same time as everyone else.
The move was a hard blow for advocates of close transatlantic ties, who want Western countries to present a united front against adversaries such as China.
The export controls on Anthropic will inevitably shadow the G7 AI discussion. Anthropic CEO Dario Amodei will attend the working lunch, giving US allies a rare chance to press their concerns directly to Trump.
Tech sovereignty agenda
"We need to establish American AI [...] as the gold standard for AI worldwide and ensure our allies are building on American technology," reads the AI Action Plan published by the Trump administration in July 2025.
The US dominates the sector — five American tech chief executives are attending the G7, against one per every other participating country. Yet it may ultimately be the US tech industry that pays the long-term price of the export restrictions.
"Tech is more and more becoming a strategic asset. Europe must be able to act on its own terms," Regnier said, calling the episode further proof that the EU must strengthen its technological sovereignty.
US President Donald Trump had already galvanised the EU's tech sovereignty agenda, persuading Europeans of the need to build domestic alternatives and reduce their strategic reliance on foreign providers, be they American or Chinese.
By locking non-Americans out of its most advanced AI models, Trump may have gone a step further: proving that Washington can shut enemies and allies alike out of the technology of its choosing, overnight.
When nightmares come true
The episode confirmed one of Brussels' longest-running fears: that American technologies carry a "kill switch" capable of disabling critical systems at will, whether fighter jets, financial institutions or critical infrastructure.
"The Anthropic kill switch shows that tech sovereignty was never abstract," MEP Brando Benifei (S&D, Italy) told Euronews.
"The G7 should not lock allies into competing AI dependencies. Europe must cooperate with the US, Canada and democratic partners, but from a position of strength."
Brussels has already mapped out its tech sovereignty agenda, but European capitals remain far from united on how far they are willing to go, or how much they are prepared to spend to unwind deeply entrenched dependencies.
Trump may have, once again, given Europe's resolve a boost, by proving that kill switches are anything but hypothetical.