The Council for Science and Technology, an independent body that advises the government, has published advice warning that Britain risks becoming what it calls a "high-value AI customer" rather than a country that shapes the technology itself. The advice was made public on 19 October, having originally been sent to the AI minister back in July.
What the advisors are warning about
The core worry is dependency. Without "deliberate and urgent action," the letter argues, the UK will end up relying entirely on AI systems built, trained and controlled elsewhere, with little say over how they evolve or what values they encode. That matters more as AI moves from chatbots into agents that take real-world actions, including in robotics.
The council's recommendations are specific. It wants a single, clearly accountable government lead for AI sovereignty rather than the current spread of responsibility across departments. It calls for sustained investment in skills, data infrastructure and compute. And it recommends pilot trials of sovereign AI capability in health, education and national security, areas where relying on foreign-controlled systems carries the most risk.
A British model?
Perhaps the most eye-catching idea is a possible publicly-governed British base model, a foundation model that researchers, startups and public services could build on without depending on US or Chinese providers. The advice is careful to say Britain should keep access to the world's best AI systems wherever they are built, while exploring domestic alternatives in parallel, not instead.
This builds on a Commons Science, Innovation and Technology Committee report from July that made similar demands. The government's response in September pointed to existing commitments: the Sovereign AI Fund, an AI Hardware Plan, and the AI Opportunities Action Plan. Whether those add up to the "deliberate and urgent action" the council wants is, by implication, still an open question.
Our take
This is a notably blunt assessment from a body that is meant to advise quietly, not campaign publicly. The gap between what the government has already announced and what its own advisors say is still needed suggests the existing plans are seen as necessary but not sufficient. For robotics specifically, sovereignty is not an abstract concern: a British-built humanoid robot or surgical system is of limited use if the AI reasoning inside it is licensed from, and ultimately controlled by, a handful of American or Chinese labs. Whether a publicly-governed British model is realistic at the scale needed to matter is a much harder question than whether it is desirable.