Britain plans to use AI to hunt for weaknesses in government systems. At the same time it says it is preparing to deal with threats posed by autonomous AI agents.
Sweeping plans announced on Monday are taking aim at both benefits and risks of AI. First, the UK government said it wants to leverage the power of AI to help the UK defend critical infrastructure and national security. Second, the government has identified the harms that powerful AI can wreak on those very systems, and so ministers are considering tougher protections against those AI threats. As part of the latter effort, the National Cyber Security Centre (NCSC) and AI Security Institute (AISI) have been tasked with figuring out how Britain should respond to destructive work from autonomous AI agents — regardless of whether it is intentional or not.
The announcement follows two government pilots that have already helped identify and address more than 1,000 potential cybersecurity vulnerabilities across public sector systems. Ministers now intend to expand the programme, using advanced AI models to identify weaknesses and detect threats before they can be exploited.
The government is also looking at whether the Cyber Security and Resilience Bill should go further to tackle threats from AI systems that can act without direct human input.
One concern is that AI systems could behave unexpectedly or slip beyond human control. That is not an entirely hypothetical issue. In June, an OpenAI AI agent (not a human but an AI-powered bot) accessed Australian government systems without authorisation, its actions triggered inadvertently after it encountered restrictions while carrying out a research task. Ministers fear similar behaviour could put essential services, such as hospitals and utilities, at risk.
Speaking at Imperial College London’s Base One technology incubator, AI Minister Kanishka Narayan said Britain needed to prepare for threats that could develop faster than existing national response mechanisms can manage.
“I want Britain to become one of the hardest countries in the world to attack using artificial intelligence,” Narayan said. “We will use every power at our disposal. We will legislate wherever necessary. And we won’t look to law alone.”
He also urged developers of advanced AI systems to investigate unexpected behaviour and avoid releasing technology where serious risks cannot be adequately controlled.
The NCSC and AISI are examining how the government should respond to major incidents involving AI agents, which can carry out complex tasks with relatively little human intervention. According to the government, such incidents could spread more quickly and affect a wider range of systems than conventional emergencies, requiring new technical expertise and response capabilities.
UK Cyber Minister Liz Lloyd said the measures would help strengthen the security of public services and improve Britain’s preparedness for emerging threats.
“Through scaling up projects piloting AI to protect public services and progressing a Cyber Bill to boost our defences, government is putting cyber security and AI preparedness at the heart of the UK’s digital future,” she said.
Yet for all the government’s warnings about AI threats, much of its response remains on the drawing board.
Ministers have yet to say when they might introduce new safeguards, what they would require of AI developers, or how Britain would respond if an autonomous system did cause a major security incident.
We will update this post as we learn more.
