TL;DR

Chancellor John Healey’s first Budget, delivered on 28 October, confirmed the AI package he had trailed beforehand and added a new £40m National Agentic Security Unit to detect and respond to AI systems that stray beyond their intended controls. City AM puts the total at more than £700m. Industry voices welcomed the safety focus but questioned whether public money will reach British suppliers.

What was new on Budget day

Most of the money had already been announced. We covered the pre-Budget package, including £120m for the AI Security Institute, the £60m National Cyber Shield and the Sovereign AI Fund, in our 26 October article.

The additions, according to City AM, were:

• £40m for a National Agentic Security Unit, which will build the capability to monitor autonomous agents that go rogue and to handle incidents involving them
• £24m for cyber security in public services, covering hospitals, schools and councils
• A consultation next year on who should carry liability for cyber breaches, and on tackling insecure technology before it reaches users
• £23m for research into AI’s economic effects

In the Commons, Healey said “nowhere is the necessary role for the state clearer than in Artificial Intelligence”, and told MPs the Sovereign AI Unit would receive “another £140m” (per the official transcript). The Treasury, citing OECD estimates, said AI could lift annual UK productivity growth by between 0.4 and 1.3 percentage points over the coming decade.

The sovereignty question

Industry responses split. Tarek Nseir of Valliance welcomed the safety emphasis but said firms need clearer guidance. Mark Boost, who runs the British cloud provider Civo, warned that if public spending “continues defaulting abroad”, the value of AI will flow to foreign platforms. Red Hat’s Jonny Williams, who advises UK public bodies on digital strategy as its chief digital adviser, said the Budget “falls short of offering the pragmatic solutions needed” for digital sovereignty, and called for open standards in procurement.

Looking Forward

The agentic unit is the measure with the most direct bearing on businesses. In our view, UK organisations deploying AI agents should expect its work to shape incident-reporting expectations, and the liability consultation could change who pays when an insecure product is breached. The details of both are still to come. The sovereignty criticism is the harder test: it will be answered by where procurement budgets are actually spent, not by the size of the funds announced.