Ministers will examine extending the Cyber Security and Resilience Bill regime to cover the identification and reporting of serious AI incidents and near misses, as part of a package of measures to protect the UK from risks posed by increasingly powerful and autonomous AI systems, the Cabinet Office has announced.
The measures, announced on 12 October alongside the AI Security Institute, the National Cyber Security Centre and the Department for Digital, Culture, Media and Sport, also commit the government to expanding the use of frontier AI models to find and fix vulnerabilities in government systems, and to developing new capabilities for responding to incidents involving AI agents.
In a speech at Imperial College London's Base One incubator, Kanishka Narayan, Minister for Artificial Intelligence, said the government would work with Parliament and technical experts to assess how the Bill's regime could be strengthened against AI-driven cyber threats and loss-of-control risks, including reporting on serious incidents and near misses and ensuring serious AI incidents can be properly reviewed.
The government said the detail of any measures would be worked through with parliamentarians, technical experts and the operators who would have to deliver them.
Narayan said the government would explore strengthening parts of the Bill and the wider regime so that essential services such as hospitals and utilities are better protected from emerging AI-related risk, including the risk of autonomous AI systems evading human oversight.
The Cyber Security and Resilience Bill will update the Network and Information Systems Regulations 2018, which impose security and incident-notification duties on operators of essential services and relevant digital service providers, with the ICO acting as competent authority for digital service providers.
Those duties sit alongside the 72-hour personal data breach notification requirement under Article 33 of the UK GDPR. Any extension of the regime to AI incidents and near misses would add a further reporting layer for public bodies and essential service operators already subject to both.
Narayan pointed to an incident in Australia earlier this year in which an OpenAI agent, tasked with finding information on public spending on medicines, encountered access restrictions on a government website and found another route in, gaining access to information it was not authorised to have. The Australian government has said no personal medical records are known to have been accessed. Narayan said the incident concerned him and that it seemed likely, if not inevitable, that the UK's own cyber defences would be breached.
Narayan said that as AI systems are used more widely with less human oversight, government must be able to establish quickly what an agent was asked to do, what it accessed, why it acted as it did and whether the same failure could occur elsewhere, which requires expertise the state has not previously needed.
The DCMS, the NCSC and the AI Security Institute will develop ways to respond to agentic AI incidents, looking in particular at how incidents can be contained and how those affected can be supported. The government said a major incident involving AI agents could play out more quickly and widely than the major incidents governments are designed to respond to.
On the defensive side, Narayan said two government pilots using frontier AI had helped identify and address more than a thousand vulnerabilities across government systems, and that the government's central vulnerability-monitoring service had cut the time taken to fix identified weaknesses from 50 days to eight.
A Government Cyber Coordination Centre case study published in June described a month-long series of hackathons in which teams across nine government organisations used frontier models to scan public code repositories, producing 407 findings including weaknesses exposing services to authentication bypass, data exposure and remote code execution, at a token cost of £13,000.
All critical weaknesses were remediated and no evidence of exploitation was found. The case study stressed that a human expert validated every finding and that AI-generated candidate findings arrive far faster than humans can triage them.
Narayan said he, the DCMS Cyber Minister and the Security Minister would lead the urgent use of frontier AI to find weaknesses in government systems before attackers exploit them. Dan Jarvis, the Security Minister, will lead a cross-government programme to strengthen defences against cyber attacks, including those using AI, following the Government Cyber Action Plan published earlier this year and ahead of a National Cyber Action Plan the government said would be launched soon.
Narayan also confirmed that the AI Security Institute is conducting evaluations of the latest generation of models, including Gemini 4 Argon, after work with the US government and leading developers to retain pre-release access. He said the government would set out support for AI verification work, which checks claims about how AI systems have been developed and tested, in the coming months.
Liz Lloyd, the Cyber Minister, said it was now business-critical for organisations of every size to be on the front foot on cyber security. "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," Lloyd said.
Narayan said he wanted Britain to become one of the hardest countries in the world to attack using artificial intelligence. Addressing frontier AI developers, he said: "Listen when your scientists warn of serious dangers. Investigate unexpected behaviour. And if you cannot adequately control the risks, do not release the system."
