US and EU discuss "trusted partner" access to advanced AI models

In short: The US and Europe are discussing a “trusted partner” plan that could give close US allies special access to the newest AI models.

What happened

US and European officials have talked about creating a “trusted partner” scheme for advanced AI systems, according to the Financial Times. The talks took place around the G7 summit in France, involving US commerce secretary Howard Lutnick and European diplomats.

The discussions come days after the Trump administration blocked Anthropic from supplying two of its latest AI models, called Mythos and Fable, to foreign customers. The US cited national security concerns. This move raised worries in Europe and in parts of the US tech industry that access to US-built AI tools could be used as political pressure.

EU tech chief Henna Virkkunen said Washington should avoid “discriminatory” steps against partners like the EU. She said the two sides should clarify what the security concerns are and how to reduce the risks.

Anthropic has said Mythos can help find serious cyber security weaknesses (like a very skilled tester looking for unlocked doors in a building). The US government ordered Anthropic to shut off access for non US nationals after a way was found to bypass the tool’s safety controls, and Anthropic said it is working with the administration to address those concerns.

OpenAI was also in the process of providing access to an advanced model to the EU’s cyber security agency Enisa and to Nato, according to people familiar with the process.

Why it matters

If countries and institutions in Europe cannot reliably access top US AI tools, it could affect cyber security testing, research, and public services that want to use them. It also adds pressure on Europe to rely less on US technology and build more of its own.

Source: Financial Times