G7 Explores 'Trusted Partner' Access to Restricted U.S. AI Models

G7 Leaders Weigh Exemptions to U.S. AI Access Ban

Leaders from the Group of Seven nations discussed creating a framework that would allow select "trusted partners" to access cutting-edge U.S. artificial intelligence models despite a recently imposed ban on foreign nationals, according to three diplomatic sources.

The conversations took place on the sidelines of the G7 summit opening dinner in Evian-les-Bains, France, where delegates raised the issue primarily with U.S. Commerce Secretary Howard Lutnick. The discussions centered on potential exemptions from Washington's decision last week to suspend foreign national access to Anthropic's most advanced AI models, a move justified on national security grounds.

According to sources familiar with the talks, the proposed "trusted partner" designation could apply to both countries and individual companies, though no formal announcement is expected when technology issues appear on the G7 agenda Wednesday.

The Anthropic Model at the Center

The restrictions have particular implications for Anthropic's Mythos model, a system designed to identify vulnerabilities in computer code. Cybersecurity experts warn that such capabilities could significantly amplify attacks on financial institutions' technology infrastructure if misused.

No European banks have received access to Mythos to date. The European Union has formally requested access to the model specifically to evaluate its security implications and potential risks.

Why it matters

This diplomatic tension reveals the growing friction between AI leadership and alliance management. The United States faces pressure to maintain technological advantages while preserving intelligence-sharing relationships with close allies. How Washington structures access to frontier AI systems—and whether it creates formal tiers of trusted partners—will set precedents for how democratic nations balance innovation security with international cooperation. For companies developing advanced AI, the outcome could determine which markets remain accessible and under what conditions.

Uncertain Path Forward

The sources, who requested anonymity due to ongoing negotiations, indicated that discussions remain preliminary. The absence of a planned statement suggests G7 members have not yet reached consensus on how such a trusted partner framework would operate or which entities would qualify.

The Financial Times first reported news of the trusted partner scheme discussions.

Details were originally reported by Reuters.