After DSIT, Britain’s next AI test is sovereignty

Britain's Department for Science, Innovation and Technology (DSIT) has been abolished, leaving uncertainty about the future of the country's AI ambitions. While the elimination of DSIT's responsibilities into other departments raises questions about coordination, the promotion of Kanishka Narayan to Minister for AI provides dedicated oversight. Recent scrutiny of public-sector AI contracts has highlighted the need for stronger control over AI operations and deployment continuity.

Britain has prioritized the speed of AI adoption, but the next step is ensuring operational control within public services. Sovereignty goes beyond data residency, encompassing an organization's ability to understand, govern, and maintain AI systems even if the provider, contract, or political relationship changes. For critical public services like healthcare, defense, and administration, data location is crucial but only part of the equation.

The key test of sovereignty is whether an organization can comprehend the system, control its usage, and continue operations if the commercial relationship shifts.

While Britain cannot replicate entire AI stacks, it must maintain predictable control over public services that rely on AI. Modern AI systems depend on international research, complex supply chains, and open-source projects, making domestic reproduction costly and potentially less effective. Instead, Britain should seek predictable control where public services rely on AI systems.

Government AI plans sometimes present sovereignty as a slower alternative to using larger commercial models. However, the focus should be on selecting the most suitable model for specific tasks rather than choosing the largest generic option. For many public workloads, a fit-for-purpose model in a controlled environment can be more valuable than a large, generic system, particularly when strict governance is required.

A specialized model tailored to an organization's data, tested under operational conditions, and governed according to decision consequences can be more useful and independent. Open and portable technology enables organizations to run models across different infrastructure, examine system behavior, and replace components with better options as they emerge. Openness does not guarantee sovereignty, but it reduces the risk of an entire service becoming dependent on a single provider's platform.

Britain possesses strong universities, experienced researchers, growing AI companies, and public investment in compute. The next step is to connect these assets with practical demand in areas where sovereignty has the most significant value. A Minister for AI can bring together different parts of government, but the strategy should be judged by whether public institutions can utilize the best technology available while retaining meaningful control over critical systems.