Macron Calls US AI Export Controls 'Strictly Nationalist'

When the US blocks its own allies from accessing its most advanced AI models, it risks causing the exact geopolitical fragmentation it was trying to prevent. That tension came to a head at the G7, where SecurityWeek reports French President Emmanuel Macron acknowledged it was a "good thing" that US officials recognized frontier AI could be dangerous, then labeled the Trump administration's export ban on Anthropic's Mythos 5 and Fable 5 models "strictly nationalist." He wasn't disputing the risk assessment; he was disputing who gets to decide who bears it.

The economic argument Macron made at a G7 working lunch - attended by OpenAI CEO Sam Altman, Google DeepMind CEO Demis Hassabis, and Anthropic CEO Dario Amodei - was pointed. If the US can turn off model access "like a light switch," it risks not just harming the European economies that have built on that infrastructure, but also triggering "a possible drop in value" for the US firms that built it. The framing recasts the export control as a business risk for Anthropic and its peers, not just a diplomatic grievance.

Macron's proposed solution was a government-to-government cooperation framework among wealthy democracies, explicitly aimed at keeping advanced AI out of authoritarian hands rather than allied ones. Altman offered a compatible pitch: an "international forum" for countries to draw up AI guardrails together, which also argues, conveniently, against unilateral restrictions that lock out partners.

The reporting doesn't spell out what a cooperative access framework would require in practice - what vetting standards, what reciprocal obligations, what happens when a US security agency objects to a specific ally. France's contingency of boosting its own AI industry funding if international cooperation breaks down suggests Macron doesn't expect easy agreement. Whether the Trump administration views these arguments as persuasion or diplomatic noise is a question the coverage leaves unanswered.

For AI companies, a structured allied-access agreement would be a considerably better outcome than the current binary. For European governments and enterprises that have already standardized on US AI systems, the impasse has made every infrastructure procurement decision more complicated in ways that will not resolve quickly.