White House restores UK AISI access to frontier AI models

The White House has restored the UK AI Security Institute’s early access to new AI models after previously blocking that access. AISI tests AI systems for safety risks before they ship, and will again be able to review models from Anthropic, Google DeepMind, and OpenAI.

The change follows a September directive from the Trump administration telling companies to stop sharing models with the British agency. The reversal reopens a channel for pre-release safety testing between major AI labs and the UK body.

According to the article, diplomatic pressure from AI Minister Kanishka Narayan led to the block being lifted. After the September move, Anthropic had restricted access to its new Claude Mythos 5.1 model.

The immediate takeaway is that access decisions for frontier AI models remain tied to government-level negotiations. For readers tracking AI policy, this is a sign that safety testing arrangements can shift quickly when national interests and diplomatic pressure collide.

Narayan plans to announce new AI safety measures in an upcoming speech, including stricter reporting requirements for AI-related security incidents involving critical infrastructure. That points to a policy focus on both pre-release model scrutiny and incident reporting after deployment.

The article also notes that the heads of MI6, MI5, and GCHQ warned ministers that AI was empowering hostile actors and terrorists. The restored AISI access therefore sits within a wider UK security discussion around advanced AI systems.

Prime Minister Andy Burnham had previously pushed for international AI regulation at the UN General Assembly. At the same event, Trump warned that “globalist” organizations should stay out of American AI development.

The restored access does not erase those tensions, but it does show that practical cooperation on AI testing can continue despite public disagreements over regulation. The key issue to watch is whether early-access safety review becomes a stable cross-border process or remains vulnerable to policy reversals.