US Reverses AI Testing Ban on UK, Restores Access to Top Models

Background

The UK's AI Security Institute (AISI) was set up after the 2023 AI safety summit. Its core job is to evaluate what advanced AI systems can do and what risks they might pose.

What happened and how fast it unraveled

On September 24, 2026, the Trump administration asked OpenAI and Anthropic to stop sharing new AI models with the UK's AISI. The hold was meant to last until a comprehensive US national security review was finished.

The pause had at least one visible side effect

Anthropic released its Claude Mythos 5.1 model exclusively to US organizations during the interim period.

What this means for AI labs, governments, and investors

For the AI labs, the episode is a reminder that voluntary arrangements are only as stable as the political conditions around them. OpenAI and Anthropic were asked to change course on short notice. Anthropic, at least, adjusted a product release to match.

Open questions remain

It is not clear whether the US national security review has concluded or what it found. It is also unclear whether Anthropic will now expand Claude Mythos 5.1 beyond US organizations, as its interim approach suggested it might.

Watch for the restored arrangement to get formalized or stay voluntary

A written framework would offer labs and allied governments more predictability.

UK AI minister Kanishka Narayan lobbied to reverse the decision

On October 11, 2026, the US reversed its earlier directive, and the UK's access was restored.

The restored access covers three of the biggest names in the field: Anthropic, Google DeepMind, and OpenAI

Notably, Google DeepMind was not named in the original September request, which targeted OpenAI and Anthropic specifically.

The episode is a reminder of the importance of AI development and testing

The episode serves as a reminder that voluntary arrangements in AI development can be fragile due to changing political conditions.