G7 weighs ‘trusted partners’ framework to expand access to advanced US AI systems

Leaders at the G7 summit are discussing a proposal that could allow select “trusted partners” to gain access to advanced artificial intelligence models developed by US companies, potentially easing restrictions imposed by Washington on foreign use of cutting-edge AI technology.

According to diplomatic sources, representatives from several G7 nations raised the issue with US Commerce Secretary Howard Lutnick during meetings on the sidelines of the summit in Evian-les-Bains, France. The discussions come days after Anthropic restricted access to its most advanced AI systems, Fable 5 and Mythos 5, following a directive from President Donald Trump citing national security concerns.

Under the proposal, access could be extended to trusted countries or companies, enabling allies to use powerful AI tools for purposes such as strengthening cybersecurity defenses and protecting critical infrastructure from sophisticated threats.

The issue is expected to feature prominently in talks with executives from Anthropic, OpenAI and Google, who are attending a G7 working session focused on AI regulation, infrastructure and emerging technologies.

Advanced models such as Anthropic’s Mythos have drawn attention from policymakers and cybersecurity experts because of their ability to identify vulnerabilities in software systems. Supporters argue that broader access among allies could improve cyber resilience, while critics warn that the same capabilities could be misused for offensive cyber operations.

Before the latest restrictions, Anthropic had provided limited access to Mythos for selected organizations in more than 15 countries, including entities in healthcare, communications, power and water services. The European Union has also expressed interest in studying the model and its implications.

The discussions highlight growing efforts by Western governments to balance national security concerns with the need for international cooperation in the rapidly evolving AI sector.