The UK government will spend £130m on AI cyber defence, including a National Cyber Shield based on agentic AI for critical infrastructure. The funding is part of an AI package of more than £600m that Chancellor John Healey will present in the Budget this week. For business, this matters because the state is moving from testing models to operational protection of infrastructure.
How the £130m cyber defence funding is structured
HM Treasury outlined the plans in a statement on Saturday, ahead of the Budget scheduled for Wednesday. The National Cyber Shield receives £60m, including a £25m challenge fund for UK AI and cybersecurity companies. The government will also scope a regulatory sandbox for agentic security tools across critical infrastructure, the public sector and industry.
Another £70m will establish a National Exposure and Real-Time Vulnerability Centre, known as NERV. It will use AI tools to identify cyber vulnerabilities across critical infrastructure and the wider economy. The statement does not specify deployment dates or the degree of autonomous action without human oversight. That gap leaves the operating model open for now.
The cyber funding sits alongside £120m for the AI Security Institute, which tests advanced AI models. A new NCSC AI Cyber Security Centre of Excellence in Manchester will train government cyber and AI specialists. The timing follows PHP2Shell flaws that took government websites across Europe offline. AI minister Kanishka Narayan said AI tools had found 500 previously unknown flaws in parts of the public sector.
What this means for companies working with AI
For large operators in energy, transport, finance and public-sector supply chains, a state-backed shield and vulnerability centre could change procurement requirements. Contracts may increasingly expect compatibility with national detection tools and documented patching processes. Small vendors gain a direct entry point through the £25m challenge fund, while large contractors face closer scrutiny of resilience.
The regulatory sandbox is the point to watch before drawing conclusions about autonomy and liability. It remains unclear what agentic security tools will be permitted to do inside live infrastructure and where human approval stays mandatory. Companies should ask vendors how agents log actions, how false positives are handled, and how tools integrate with existing monitoring. This announcement alone does not guarantee deployment timelines.
The marker to follow is the Budget presentation on Wednesday and the subsequent terms of the challenge fund and sandbox. If eligibility, funding rounds and pilot sectors appear, the programme shifts from declaration to procurement. Former National Cyber Security Centre head Ciaran Martin framed the task as locking down systems as AI exposes weaknesses and creates new attack methods.
