UK must build sovereign AI, warns Alan Turing Institute head

Britain must stop relying on foreign artificial intelligence systems that could be "switched off at short notice," the head of the Alan Turing Institute warned on October 10.

George Williamson, who runs ATI, said the UK's response to "runaway leadership" by the United States and China in AI must centre on "national resilience." The intervention reframes a research agenda as a sovereignty question.

"The country needs a stronger domestic position in AI in response to the US and China's runaway leadership in the field," Williamson said.

What does Williamson mean by "national resilience"?

He named "national resilience" as a key focus for the institute. The phrase signals a pivot toward domestically controlled AI capability — models, compute and infrastructure housed inside UK jurisdiction.

The argument is operational as much as political. AI services delivered by overseas providers sit beyond British legal reach. Licensing terms, sanctions regimes or a simple corporate decision can cut access. A model trained in California or Shenzhen can be denied to a London ministry or hospital with little warning and less recourse.

Why does he call US and China "runaway" leaders?

Williamson characterised the gap with Washington and Beijing as one of pace and scale. The two ecosystems command the deepest pools of capital, the largest captive chip supply and the biggest domestic markets for their own output.

Britain, by contrast, has strong academic talent but a thinner industrial base. The risk Williamson flags is structural. Without domestic capability, British institutions end up renting AI from the same two blocs they cannot fully influence.

What is ATI's role in this strategy?

The Alan Turing Institute operates as Britain's national centre for data science and AI. It coordinates research across universities and advises government on AI policy. Williamson's framing therefore carries weight: he sets the institute's strategic direction.

By making "national resilience" the headline focus, he signalled ATI will tilt toward applied work with direct strategic value. That means projects that build domestic capability — local models, local datasets, local compute — rather than pure academic publication.

What is the kill-switch risk?

"Switched off at short notice" is the phrase Williamson used to describe the threat. It captures a set of operational and political disruptions that range from licence changes to sanctions to a foreign government's intervention in a private supplier.

The risk lands hardest in sectors Britain cannot afford to see blacked out — defence, health, critical infrastructure. Losing AI access in any of those would have immediate operational consequences.

What happens next?

The warning lands as London weighs its place between the US and Chinese AI stacks. The cheaper path — buying compute and frontier models from US hyperscalers — is exactly the path that triggers Williamson's concern.

The more expensive path is sovereign AI: building domestic capacity in compute, chips and models. That route demands capital, talent and time Britain has not yet committed at scale.

The country's choice, in Williamson's framing, is between paying for sovereign capability or accepting a dependency it cannot switch off itself.