G7 governments are discussing a way to share cutting-edge U.S. AI models with allied countries under a “trusted-partner” framework, according to Reuters.

The talks, held among G7 leaders, are aimed at figuring out how access could work when the systems are considered sensitive. Rather than treating all users the same, the approach being discussed would link model availability to partner status and safeguards.

Reuters reports that the discussion included access to models from companies such as Anthropic. That matters because it points to a practical, multi-firm path rather than a single “one door” deal, even though the details of how that would be handled were not set out in the reporting.

Why the trusted-partner idea is on the table

What’s driving the conversation is the security and defense angle. Reuters said the proposal was framed around allied access to frontier models for security and defense use cases. In everyday terms, that means governments want a way to work with advanced AI while limiting who gets in, and under what conditions.

Officials are likely trying to balance two competing needs: keeping high-end models within trusted circles, while still allowing allies to use them for urgent tasks. If the framework moves forward, it could shape how countries cooperate on AI capabilities, and how companies are asked to support access rules for governments and defense-linked projects.

For now, Reuters described the effort as a weighing of options. No final structure was reported, so it’s still unclear how broad access would be, what verification would look like, or how exactly security requirements would be enforced.

Source: Reuters