UK announces AI risks package and extends cyber defense pilots – Unite.AI

On 12 October 2026 the UK Government announced a package of measures to protect Britain from the risks posed by increasingly powerful artificial intelligence, including the extension of a Frontier AI Cyber ​​Defense program which the Government says has already helped find and fix more than 1,000 potential vulnerabilities in government systems.

This was announced by the Presidency of the Council of Ministers, the AI ​​Security Institute (AISI), the National Center for Cyber ​​Security (NCSC) and the Department for Digital, Culture, Media and Sport. AI Minister Kanishka Narayan outlined the measures at Base One, the science and technology incubator at Imperial College London, arguing for a whole-of-society approach to resilience. He warned that as frontier AI systems become more powerful and capable of acting autonomously, the potential consequences of failure or malicious use could become much more severe.

“I want Britain to become one of the hardest countries in the world to attack using artificial intelligence,” Narayan said. He pledged to legislate where necessary and to rely on Britain’s scientific expertise, technological capacity and institutions to understand increasingly powerful systems, counter emerging threats and respond when things go wrong. He said companies building the world’s most powerful artificial intelligence systems should heed their scientists’ warnings of serious dangers, investigate unexpected behavior and block any systems whose risks cannot be adequately controlled.

According to the announcement, the initiative is based on two pilot projects that helped identify and fix vulnerabilities and detect potential cyber threats. Extending the programme, the Government said, will help protect public services that people rely on every day, reduce short-term risks and support long-term take-up across the public sector.

Changes to cyber invoices and preparedness for AI incidents

The Government has said it will explore strengthening parts of the Cybersecurity and Resilience Bill and the wider regime so that essential services such as hospitals and public services are better protected from emerging AI-related risks, including the risk of AI loss of control, where autonomous systems escape human oversight. It will also look at how to identify and report serious AI-related incidents and near-misses, and how to quickly learn lessons to reduce the risk of future harm, with details of any measures agreed with parliamentarians, technical experts and operators who would need to implement them.

Work is also underway with the NCSC and AISI on how to respond to incidents involving AI agents, systems capable of carrying out complex tasks with little human supervision. Such an incident could unfold faster and spread more widely than other major incidents that governments are called upon to manage, the announcement said, and meeting this challenge will require new skills and capabilities.

The package also extends to AI testing, which the announcement describes as an emerging field that tests whether claims about how AI systems were developed, tested and evaluated stand up to scrutiny. The Government has said that the UK is well placed to take a leading role in this field and that more details on supporting AI testing efforts in the UK will be set out in the coming months.

IT Minister Liz Lloyd said cyber incidents continue to disrupt organizations large and small across the public and private sectors, creating what she described as a critical need for businesses to remain at the forefront of cyber security. “Through expanding projects piloting AI to protect public services and bringing forward cyber legislation to strengthen our defences, the Government is putting cyber security and AI readiness at the heart of the UK’s digital future,” Lloyd said.

The driver behind the vulnerability figures

The announcement cites a case study published on 12 June 2026 by the Department of Science, Innovation and Technology (DSIT) and the NCSC, to support its figures. It documents a pilot project in which the Government Cyber ​​Coordination Center (GC3), a partnership between NCSC and DSIT, held weekly in-person hackathons that used frontier AI to scan public code repositories across the government. Working with AISI and NCSC specialists, the teams sought to identify and mitigate previously unidentified vulnerabilities before they could be exploited. The CG3 did not impose a single approach; teams received access to the model and created their own tools.

The case study describes several pipelines. A team ran each public repository through a chain of AI agents in six stages (triage, validator, auditor, tracer, judge, and summarize) with each stage reading and questioning the last, with manual escalation: a team member checked each row, rechecked for exposure, and handled false positives. Another team first used conventional scanning tools, including Gitleaks, Trivy, Semgrep, and Hadolint, then overlaid three model phases covering discovery, attack chain investigation, and triage. A third department distilled an organization-wide audit of hundreds of services into five domain-specific Claude competencies.

Participants identified 407 findings in total, including critical weaknesses that expose services to authentication bypass, data exposure, and remote code execution, the case study reports. All critical weaknesses were addressed and no evidence of exploitation was identified for any outcome, costing £13,000 in tokens across nine government organizations over the month. The authors said the models tracked vulnerabilities across service boundaries, which traditional scanners can’t do, and linked business logic with technical details.

One critical discovery affected legacy GitHub stocks in a repository supporting a key government digital service. A specially crafted comment on an open pull request could allow an external user to initiate a chain of workflows, resulting in arbitrary remote code execution on the GitHub Actions runner and a potential path to extract available secrets and tokens for that workflow. GC3 will launch a second phase of the pilot project with more departments, additional models, and an extension from public code to closed-source systems.

Evaluations of models and the broader government program

Narayan confirmed that AISI scientists are conducting rigorous evaluations of the latest generation of advanced AI models, following work with leading AI developers and the US government.

The announcement also refers to an open letter to business leaders published on 15 April 2026, in which DSIT Secretary of State Liz Kendall and Security Minister Dan Jarvis urged businesses to take cyber security seriously at the top of their organisations, get the basics right with the Government’s Cyber Essentials scheme and follow the NCSC’s advice, including signing up to its free early warning service, which alerts organizations to potential cyber attacks so they can act before a accident escalates.

According to the announcement, the measures are the first steps of a longer journey and broader work is being led by the Prime Minister’s Task Force on Artificial Intelligence. The announcement highlights the Prime Minister’s commitment to the United Nations General Assembly in September to do everything necessary to protect the British people from AI-related risks as they arise, and says that AI will be at the heart of the UK’s G20 presidency. In the coming months, the task force will focus on the areas most impacted by AI, ranging from national security, economic growth and the future of public services and the workforce.