I am today updating the House on the action the Government is taking to strengthen the UK's resilience against the risks posed by increasingly powerful artificial intelligence.

Artificial intelligence is becoming increasingly capable and autonomous, creating new risks for individuals, businesses and national security. Deepfakes are being used to defraud people out of their money. Nudification tools are being used to humiliate and blackmail them. And AI enabled cyber attacks threaten our businesses and public institutions.

As I set out to the House on 7 September, frontier AI agents are increasingly able to act on their own. As these systems become more powerful, we must also prepare for the possibility of serious incidents in which they behave unexpectedly or act beyond their instructions.

The Government's first responsibility is to protect the British public. That requires the independent expertise to understand the risks posed by advanced AI, the capabilities to strengthen our defences and the ability to respond effectively when things go wrong. We cannot rely solely on the companies developing these technologies to assess and manage the risks they may pose.

I am therefore setting out further action to strengthen government cyber defences using frontier AI, establish a specialist capability to respond to major incidents involving agentic AI, and consider how the Cyber Security and Resilience Bill can better protect essential services against emerging threats.

As artificial intelligence becomes more capable, Britain must become more capable of protecting its people.

Independent testing of advanced AI

Britain is home to the world's leading AI Security Institute, bringing together more than 100 technical specialists to independently test the capabilities and risks of the most advanced AI systems.

Its work has established the UK as a world leader in AI security and demonstrated why independent scrutiny matters. As these systems become more powerful, it is essential that government retains the expertise to understand their risks for itself, rather than relying solely on the companies developing them.

So, over the last few weeks, we’ve been working closely with the labs and the US government - from His Majesty’s gathering of leading AI developers at Dumfries House in Scotland to discussions at the United Nations General Assembly in New York - to retain access for Britain’s AI Security Institute to the most powerful AI models before they’re released. And I can confirm that AISI scientists are now conducting rigorous evaluations of the latest generation of models, including Gemini 4 Argon to identify risks early, test safeguards, and help keep people safe.

Frontier AI hardening across government

Over the past months, experts in the Government Cyber Unit have been working across the public sector to test the use of frontier AI on real-world government cyber security problems. Two pilot programmes have used AI to find and address 1,352 vulnerabilities across 548 code repositories in 18 participating departments, alongside helping organisations detect advanced cyber threats. I can today announce an urgent extension of our frontier AI hardening programme, putting the most advanced AI technology to work strengthening defences across government and making our vital public services harder to attack. We will continue to work with government departments to address weaknesses in highest risk systems. This marks an important step towards a broader UK Government capability for use of AI to support cyber defence and resilience.

Agentic AI incident response

As AI agents are used more widely and with less human oversight, the risk of serious incidents in which they act beyond their instructions is growing. When such incidents occur, government must be able to establish quickly what has gone wrong, contain the risks and support those affected. I can confirm today that the Government is taking the first steps to meet this challenge. Working with the National Cyber Security Centre, Department for Digital, Culture, Media and Sport and the AI Security Institute, we are developing our plans for how government should respond when AI agents cause serious harm, and we will build on this work in the weeks and months ahead.

Cyber Security and Resilience Bill

The Cyber Security and Resilience Bill currently making its way through the House of Lords will do a great deal to protect our essential services, such as in health, energy and drinking water. Given how fast AI capabilities are advancing, we are considering how the regime can be strengthened so that AI cyber and loss of control risks are better understood and protected against. For instance, how near misses can be identified and reported, how AI-driven risk across organisations is reduced, and looking at how AI incidents can be properly understood, so lessons are shared and future harm prevented. We will consider proposals with parliamentarians, as well as technical experts and the operators who would need to deliver them, and continue to update Parliament during the passage of the Bill.

Verification

As AI systems become more capable, effective risk mitigation may increasingly depend on whether important claims about their development and operation can be independently verified. The UK has a rich history of hardware innovation, and I can confirm today that we will explore how that British expertise can help strengthen this central challenge in AI security.

The Prime Minister's AI Taskforce

These are just the first steps on a longer journey Britain needs to take to seize AI's potential as a driver for good growth and better public services, while at the same time being resilient to its potential risks. That wider scheme of work is being spearheaded by the AI Taskforce, reporting directly to the Prime Minister and chaired by Lord Vallance.

A shared responsibility

Government cannot do this alone. AI is increasing the speed, scale and efficiency of attacks, and every business has a responsibility to stay secure against them. The damage to brands, reputations and revenues means cyber security must be a matter for every boardroom in Britain.

The Security Minister will be outlining this Government’s approach to cyber security for the age of AI in the upcoming National Cyber Action Plan, with a national call to action for all organisations and businesses on this shared responsibility.

AI will be as consequential as the steam engine or internet. It could be more consequential than those technologies. It can help those who seek to do us harm, but it can also help keep us safe. This Government is determined that the UK stays at the forefront of AI innovation, so that we can seize the opportunities it brings and protect the public from its risks. I will keep the House updated.

https://www.theyworkforyou.com/wms/?id=2026-10-12.hcws357.0