At the Wednesday G7 summit, several leaders — including Emmanuel Macron of France and Narendra Modi of India — voiced concern that the United States could unilaterally revoke other countries’ access to leading American artificial intelligence models. The discussion intensified after the Trump administration blocked exports of Anthropic’s latest models, Mythos 5 and Fable 5, citing national security reasons.
Macron told attendees at a working lunch that if the United States can “shut off AI access overnight,” such an action could harm not only European customers’ economies but also the AI companies themselves. According to reports, the meeting included senior figures from major AI firms, among them Dario Amodei and Sam Altman; the reporting also mentions Donald Trump by name in the list of mentioned individuals.
The export ban followed warnings from Amazon to the White House that certain security safeguards might be circumventable. Since the decision, Anthropic’s affected models have remained unavailable for export. Cybersecurity experts have said, however, that the capabilities the government cited can also be found in other models that remain accessible, including systems from OpenAI.
The episode highlighted a longstanding risk: any company or government that builds on US AI infrastructure or services must reckon with the possibility that access could be cut off for political or national security reasons on short notice, potentially without detailed explanation.
The Financial Times reported that Narendra Modi also expressed concern about the blocking of Anthropic’s models, arguing that democratic countries need uninterrupted access to the most advanced AI models to protect critical infrastructure.
Aidan Gomez, co-founder and CEO of Canadian corporate AI developer Cohere, commented on the issue in a statement. He said: “The recent restriction of access to Anthropic’s models reinforces what we at Cohere have long known: it creates a dangerous situation when companies and democratic countries are dependent on a narrow set of tech giants.” He added that digital sovereignty is not just about market competition but about who controls the foundational technologies that will determine economic security and national sovereignty in the years ahead.
At the summit, G7 leaders also discussed creating a so-called “trusted partners” framework. The idea would be to enable countries and companies outside the United States to access advanced AI models from firms such as Anthropic or OpenAI in ways less directly subject to US control. The proposal would allow not only states but also companies to qualify as trusted partners if they use the models to develop defensive capabilities that strengthen their position against rivals, including China.
However, many details remain unclear: it is not yet known how broadly such a system would be applied or whether it would offer meaningful protection for smaller actors — for example, a startup in Paris or Bangalore whose product suddenly becomes unusable because access is restricted. Macron said it would be logical for Washington to support the initiative and to widen access to Mythos models, adding that no one wants to pay for an American AI service that can become unavailable overnight.
The debate also underscored a tension: while Europe and other non-US countries are increasingly pushing for AI sovereignty, their arguments are weakened as US models consolidate advantages, and few want to be left out of the next phase of the AI revolution.
Summary
Concerns raised at the G7 summit highlight the fragility of a global AI ecosystem reliant on US-based models. The Trump administration’s export ban on Anthropic models has prompted discussion of alternative access mechanisms, but questions remain about how effective such measures would be for smaller countries and companies.
