Budget week: the UK commits £600M+ to AI security, biosecurity and sovereign capability

Ahead of this week’s Budget, the Chancellor, John Healey, has set out a package of more than £600M in AI funding — the government’s biggest single statement of intent on artificial intelligence to date, and a direct, if partial, answer to the warning its own science advisers issued this month that the UK risks becoming a “high-value AI customer.” It spans security, biosecurity, sovereign capability and skills.

What is in it

The headline allocations:

• AI Security Institute — £120M. A substantial commitment to the body that evaluates frontier models, and a notable vote of confidence after its access to US models was briefly switched off last month.
• National AI Biosecurity Initiative — £165M, comprising a £100M AI Biosecurity Lab to detect biological threats and £65M for biological surveillance and countermeasures. OpenAI and Anthropic are separately offering up to £200M in model credits to support it.
• Sovereign AI Fund top-up — £60M, with its remit widened to physical and industrial AI (robotics, autonomous systems, advanced manufacturing), plus a £60M Sovereign AI R&D Procurement Scheme — building on the procurement programme we covered.
• National Cyber Shield — £60M (including a £25M challenge fund) using AI agents for cyber defence, a £70M National Exposure and Real-Time Vulnerability Centre, plus a regulatory sandbox for autonomous security tools and an NCSC centre of excellence in Manchester.
• UK Advanced Chip Design Academy — £18M, with chip-design capability planned in Edinburgh — a small but pointed bet on domestic compute skills.
• Plus £30M for local AI partnerships, a £23M AI Economics Institute, £10M a year on AI’s economic impact, and a £15M AI Challenge Prize focused initially on health.

The sovereignty logic — and the catch

This is, in large part, the government doing what the Council for Science and Technology told it to do: invest in domestic capability — chips, sovereign compute, safety institutions — rather than rely wholly on others. The £120M for the AI Security Institute and the chip-design academy are “creator, not customer” money. The widened Sovereign AI Fund is the state putting capital behind British physical-AI companies.

And yet the flagship biosecurity lab — the single largest line — leans on up to £200M of compute credits from OpenAI and Anthropic, two American companies. It is a neat illustration of the bind: the UK is funding the institutions and the research, while the raw capability underneath still comes, in good part, from US labs — the same pattern behind Nvidia’s £189M compute pledge and the model-access freeze. That is not a criticism of taking the credits; it is the reality the sovereignty debate keeps running into.

Why it matters

Two caveats temper the number. First, this is a pre-Budget announcement, and announcements are not delivery — several of these are funds and institutes that have to be stood up before they do anything. Second, it is conspicuously quiet on the constraint that derailed the UK’s largest planned supercomputer this month: electricity and grid connections. You can fund a chip-design academy and a biosecurity lab; you still have to power them.

Still, as a statement of priorities it is coherent and substantial, and it maps closely onto the themes this publication has been tracking all autumn — safety, sovereignty, compute and the uncomfortable dependence on American infrastructure. The Budget itself will show how much of it survives contact with the public finances.