TL;DR

AI minister Kanishka Narayan is due to give a speech in west London on Monday 12 October urging the companies behind the most capable AI systems to raise their safety game. According to The Guardian’s preview, he will argue that Britain cannot leave those firms wholly in charge of policing their own models, announce steps to strengthen national cyber defences, and make the case for more home-grown AI infrastructure.

A first outing for the cabinet-level brief

It is Narayan’s first speech since Prime Minister Andy Burnham elevated AI to a cabinet-level portfolio. The preview does not detail the cyber measures themselves, so their scope will only be clear once he has spoken.

The backdrop is a run of warnings. Geoffrey Irving, who was formerly chief scientist at the AI Security Institute (AISI), told The Guardian: “The AI companies do not have security that would prevent an adversarial country from stealing the models.” He pointed to an OpenAI bug-bounty exercise in which researchers said the potential reach of their access was “huge”. OpenAI says its models are guarded by several protections, among them encryption and round-the-clock monitoring.

The intelligence agencies have also weighed in. Last week the cabinet heard from the heads of GCHQ, MI6 and MI5 about the threat AI poses to national security, The Guardian reports, citing the Sunday Times.

The international question

Observers will be listening for any hint that the UK might back an international call for coordinated safety standards. Leaders from countries including France, Germany, Ireland, Finland, Australia, Singapore and the UAE have signed it; neither Britain nor the US has, and President Trump wants the leading labs to “self-police”. At the UN last month, Burnham said the UK’s G20 presidency next year would push for “a single set of global principles and standards”.

Looking forward

The expected message shifts the emphasis. At Labour’s conference in Liverpool, POLITICO reported, Narayan cautioned that regulation may not be Britain’s most effective lever and put building British capability first. Monday’s speech keeps the capability argument but adds a plainer warning that voluntary safety falls short. In our view, the gap to watch is between that rhetoric and any means of enforcing it. The Guardian reported last month that Starmer’s ministers had begun drawing up an AI safety law, including whether labs could be forced to submit models for testing before launch, before Burnham scrapped the department leading the work. Nothing in the preview says whether this speech revives that idea. UK businesses building on frontier models should watch for any move from voluntary commitments towards compulsory reporting or testing.