THE British government has unveiled a broad set of AI security initiatives aimed at strengthening protection against emerging threats. Speaking at Imperial College London on 12 October, AI minister Kanishka Narayan outlined plans to employ frontier models to discover vulnerabilities in government systems, noting that the central vulnerability-monitoring service had already halved fix times—from 50 days to eight.
The administration said the DCMS cyber minister, the security minister and Narayan would lead a co-ordinated effort to use frontier AI to identify weaknesses before attackers exploit them, safeguard essential government systems and respond more effectively to new threats.
The programme also involves closer collaboration between the Department for Digital, Culture, Media and Sport (DCMS), the National Cyber Security Centre (NCSC) and the AI Security Institute (AISI) to develop robust responses to agentic AI incidents. Narayan cited past internal hackathons run by the Government Cyber Coordination Centre (GC3) in which frontier models uncovered more than 400 software flaws in public sector IT, all subsequently patched.
The government signalled forthcoming amendments to the Cyber Security and Resilience Bill (CSRB) to address loss-of-control risks and AI-powered threats, including incident reporting requirements and ensuring serious AI incidents can be properly reviewed.
AISI testing of frontier models is resuming, with models such as Gemini 4 Argon being evaluated to inform developers and international risk awareness. The government also floated potential enhancements like kill switches and a broader cross‑government programme to bolster resilience, while industry voices urged practical safeguards and timely incident reporting to deter attackers and maintain UK competitiveness.