The United States and European governments are discussing a “trusted partner” scheme that would let vetted allies regain access to the most capable American AI models, days after an export-control order cut off foreign access to them entirely. The talks were held largely with Commerce Secretary Howard Lutnick on the sidelines of the G7 summit in France.

The restriction the scheme is meant to address took effect on June 12, when the government ordered access to the most powerful frontier models suspended for foreign nationals on national-security grounds. Because the companies could not reliably separate foreign users from domestic ones, the order forced Anthropic to disable its Fable 5 and Mythos 5 models worldwide, including for paying customers in the United States. Allied governments lost access alongside everyone else.

The proposed arrangement would create a path back. Under the concept discussed at the summit, vetted entities from allied countries could regain access to cutting-edge models from companies such as Anthropic, OpenAI, and Google, and officials said the scheme could be defined around either countries or companies. Proponents argued that wider allied access would help G7 nations build stronger cybersecurity defenses against rivals such as China, the same capability concern that drove the original restriction. The labs whose models are at issue were on hand: a summit working session included the chief executives of OpenAI, Google DeepMind, Anthropic, and Mistral.

The mechanism extends a concept the administration has already written into domestic policy. The June executive order on frontier AI established a framework in which the government identifies “trusted partners” to receive early access to covered models and uses a classified process, run through the National Security Agency, to decide which models qualify. The summit discussions apply that same vocabulary across borders, turning access to American models into something allied governments negotiate for rather than assume.

That access would remain a US instrument. “Trusted partner” status would be defined in Washington and revocable there, leaving allied governments dependent on a designation they do not control. The European Union is separately seeking access to Mythos, the cyber-capable model Anthropic had restricted before the government suspended it, in order to study its security implications. The two tracks point in the same direction: European governments asking, through different doors, for entry to models that American firms build and American policy gates.

The episode marks a shift in what frontier-model access is. For most of the past two years, the constraint on who could use the leading models was commercial, a matter of price and availability. It is now becoming a matter of alliance standing, managed country by country in the way export-controlled defense and dual-use technology already is. A model that was globally available one week became, the next, something a government in Berlin or Tokyo may need Washington’s permission to use.

Sources: The Japan Times, Seeking Alpha

–
By the Control Plane Editorial Team