Macron, Modi warn G7 over US 'off switch' on Anthropic AI

A US block on exports of Anthropic's newest models, reportedly named Mythos 5 and Fable 5, has turned this week's G7 lunch into an unusually pointed conversation about who actually controls American AI. TechCrunch reported that French President Emmanuel Macron warned the room, which included Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman and President Donald Trump, that if Washington "from one day to the next can turn off the switch," the damage lands not just on European customers but on the AI firms themselves. Indian Prime Minister Narendra Modi raised the same concern, arguing democratic nations need unfettered access to top models to protect critical infrastructure.

The trigger, according to the reporting, was the Trump administration's decision to prohibit Anthropic from exporting the two new models on national security grounds, after Amazon flagged that certain safety guardrails could be bypassed. Cybersecurity experts cited in the piece pushed back, arguing similar capabilities already exist in freely available competing models, which makes the export logic harder to defend on technical grounds alone.

The proposed off-ramp is a "trusted partners" scheme G7 leaders discussed at the same lunch, which would let approved non-US governments and companies keep accessing advanced models from firms like Anthropic and OpenAI, provided they use them to develop stronger defenses against rivals like China. Aidan Gomez of Canadian lab Cohere framed the broader stakes in language familiar from the cloud sovereignty debates: "Digital sovereignty is not just about market competition or any one company or nation. It's about who controls the foundational technology that will shape our economic security and national sovereignty for decades to come." Al Jazeera reported Canadian Prime Minister Carney telling allies they "will have done something wrong if we just accept this, don't take the lesson, don't build out and diversify."

The honest caveat is that the reporting does not lay out what the "trusted partners" scheme would actually look like in practice, how it would treat smaller startups rather than national governments, or what Amazon specifically saw in the models that triggered the export call. Take the model-name specifics as reported, not settled.

What matters going forward is that the ban has handed non-US labs, Cohere very much included, the cleanest sovereignty pitch they have had, and given every European and Asian buyer a reason to keep an open-weight fallback warm. If even the G7's most committed allies are now publicly modeling a US off-switch, the next round of enterprise AI procurement is going to read very differently.