Sovereign by Design
AI sovereignty means the UK having the practical ability to choose, govern and sustain the AI capabilities its security, economy and public services rely on. Achieving it calls for domestic strength in selected areas, credible alternatives where overseas providers dominate, and demonstrable control over essential systems; it does not require every chip, model and application to be made in Britain.
Drawing on a high-level UKAI roundtable, this report asks a simple question: could an external actor stop the UK from using an essential AI service? The answer depends on far more than where a data centre sits. Ownership, licences, cloud administration, software updates, energy, skills and legal jurisdiction all matter.
Key findings
- The UK starts from strength. Nearly 5,900 AI firms generate £23.9bn in revenue, and public AI compute has grown tenfold in a year. But economic momentum is not the same as control over infrastructure, capital and supply chains.
- Coordination is the biggest gap. Compute, skills, data and procurement sit in separate policy tracks. The result is a collection of useful policies rather than a single strategy.
- A four-part test. The report proposes assessing sovereignty through ownership, autonomy, control and resilience, applied in proportion to the consequences of failure.
- Where the UK can lead. Fine-tuned and specialist models, Green AI, AI safety and assurance standards, and applications for regulated sectors such as law, finance, health and defence.
- Assurance depends on access. The UK’s ability to evaluate frontier models increasingly rests on access it does not control.
Headline recommendation
Government should name a single accountable owner for sovereign AI, with authority over cross-departmental procurement, backed by a common national definition and proportionate sovereignty scoring in public contracts.