Burnham takes an AI security partnership to the UN

What London is taking to the UN

Over the next 24 hours, world leaders will gather in New York while Burnham discusses:

The UK and US plan to work together on AI and autonomous systems through Britain’s Rapid AI Delivery Taskforce at the Ministry of Defence and the US Chief Digital and Artificial Intelligence Office at the Ministry of War.

The stated purpose is defensive: identify threats and deter adversaries using the latest AI technologies developed on both sides of the Atlantic. Burnham is expected to argue at the UN General Assembly on Tuesday evening that AI can contribute to a more stable and secure world.

The political setting is more delicate. This is Burnham’s first meeting with Trump, while London is also positioning AI as a central issue for the UK’s G20 presidency in 2027.

The gap in the safety message

The announcement is specific about where AI might be deployed, but vague about how advanced systems should be tested before deployment.

Sam Coates of Sky News reported that many leading countries and the European Union had backed Finland’s call for mandatory testing of all frontier AI models before release. Britain did not join them.

The Finnish initiative was published on September 21. According to Coates, it was supported by the leaders of:

The initiative is available here: https://presidentti.fi/en/a-call-for-control-of-frontier-ai-models/.

Coates said Britain was expected to be invited to participate. Burnham did not sign the call. Instead, he is expected to present the UK as a country well placed to lead international AI regulation and mediate between the US and China.

The British release does not mention support for mandatory pre-release testing. Its language is notably less concrete: AI should be used “for good,” but the government does not say which measures it wants to put in place.

That contrast gives the trip its real tension. The release reads more like a plan to apply AI to national security than a plan to control the systems before they become dangerous. I think the omission matters more than the promise of leadership: Britain is offering itself as a bridge between major powers without publicly backing the specific safety measure its likely partners are asking for.

A call to sign Finland’s proposal

Darren Jones, who served as chief secretary to the prime minister when Keir Starmer was at Downing Street, said the government should not ignore Finland’s initiative.

He praised @alexstubb for bringing countries together around the demand to test AI before deployment. In Jones’s view, every signatory to the “Call for Control of Frontier AI Models” is a natural ally of the UK.

Jones also pointed to Britain’s AI Security Institute, which gives the country special access to US AI models. That access, he said, is why the British government wants to act in sync with the US.

But Jones sees no reason Britain could not join Finland’s proposal. He urged the government to sign the document and offer the US help in reaching agreement on how it should be applied.

What remains unresolved is whether Britain’s promised partnership will include meaningful constraints before frontier models are released, or mainly make the technology easier to deploy once governments have chosen where they want it.