How Did Diplomatic Pressure Restore UK AI Safety Institute Access to US Models?

The White House has reversed a policy that prevented the UK’s AI Security Institute (AISI) from accessing new artificial intelligence models. According to The Decoder, the agency once again has early access to systems developed by Anthropic, Google DeepMind, and OpenAI. This change allows the British body to resume its safety testing protocols before these models are released to the public.

What triggered the initial access block?

In September, the Trump administration instructed AI companies to cease sharing their models with the AISI. The agency’s mandate involves testing AI systems for safety risks prior to their commercial release. Following this directive, Anthropic restricted access to its new Claude Mythos 5.1 model for the institute. This move followed the administration's order to stop sharing models with the British agency.

How was the access restored?

Diplomatic pressure from AI Minister Kanishka Narayan led to the reversal of the block. The Daily Telegraph reports that the White House lifted the restrictions after these interventions. The reinstatement of access means that the AISI can once again examine frontier models from the named companies. This development restores a key channel for international safety coordination between US developers and British regulators.

What is the broader political context?

Prime Minister Andy Burnham had previously advocated 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. Intelligence heads from MI6, MI5, and GCHQ had also cautioned ministers that AI technology was empowering hostile actors and terrorists. These statements reflect the differing perspectives on AI governance present during the period of the access block.

What comes next for AI safety measures?

AI Minister Kanishka Narayan plans to announce new AI safety measures in an upcoming speech. These measures will include stricter reporting requirements for AI-related security incidents involving critical infrastructure. The restoration of model access coincides with these planned announcements. Developers and regulators will now operate under a renewed framework for pre-release testing and incident reporting, aiming to mitigate risks associated with advanced AI systems.

What to do and how to stay safe

• Monitor official announcements from the UK AI Security Institute for updates on safety testing results and new regulatory requirements.
• Review internal incident reporting procedures to ensure compliance with upcoming stricter requirements for AI-related security events involving critical infrastructure.
• Stay informed about policy changes affecting cross-border AI safety cooperation, as these can impact model release schedules and access for external auditors.
• Evaluate current AI safety testing protocols to align with the standards being applied by the AISI to frontier models from major US developers.

Step-by-step guide: How AI Risk Management Works: The Hidden Failure Points

General security guidance from the Sandbox Dispatch newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Which companies had their model access restored to the UK agency?

Access was restored for Anthropic, Google DeepMind, and OpenAI, allowing the UK AI Security Institute to test their frontier models again.

When did the US block the UK agency's access?

The Trump administration ordered the block in September, instructing companies to stop sharing their models with the AISI.