Oxford Martin School researchers told MPs this week that the UK cannot rely on continued access to the world's most advanced artificial intelligence systems, presenting new evidence to Parliament as AI sovereignty moves up the pollical agenda.
The briefing came one day before the House of Commons Science, Innovation and Technology Committee published a report warning that the UK "may not be able to count even on its allies for access to technologies" and calling for a clearer strategy on technology sovereignty. The committee's conclusions closely echo the Oxford Martin School research presented to parliamentarians during Evidence Week.
The UK hosts significant AI infrastructure, but it does not fully control enough of it to depend on when it matters most.
The researchers identify two strategic vulnerabilities.
The first is compute, the chips, servers and data centres used to train and run advanced AI systems. The UK hosts significant computing capacity, but much of it is owned or controlled by overseas companies. Their analysis shows that while the UK hosts around 4% of the world's AI computing capacity, it exercises sovereign control over only around 0.1% of it. In practice, that means access to computing power on UK soil could still depend on decisions made outside the UK.
The second is access to the most advanced AI systems, often known as frontier AI. These models are overwhelmingly developed and hosted overseas, leaving the UK dependent on commercial relationships that can be altered or withdrawn.
Professor Robert Trager speaks with parliamentarians, policymakers and fellow researchers at Evidence Week 2026
Recent events have highlighted how rapidly access can change. US export controls temporarily forced Anthropic to suspend access to its newest AI models before the restrictions were partially lifted, illustrating how decisions taken outside the UK can quickly affect the availability of frontier AI. Alongside growing warnings from governments and the United Nations about the risks posed by advanced AI, the episode has intensified debate over whether access to these systems should be a matter of national resilience rather than simply commercial supply.
Professor Robert Trager, Director of the Oxford Martin AI Governance Initiative, said:
"Artificial intelligence is becoming a core national capability, underpinning economic, societal and security processes. The question for Parliament is whether the United Kingdom has reliable, sovereign access to the frontier of artificial intelligence."
Associate Professor Amro Awad said the underlying issue was not simply where AI infrastructure is located, but who controls it.
"We found that the UK hosts substantial compute infrastructure on its own soil. But that infrastructure is overwhelmingly owned and operated by foreign cloud providers through what is called a control plane. If those companies, or their governments, decided to restrict access, the physical servers in UK data centres could become unusable."
"If those companies, or their governments, decided to restrict access, the physical servers in UK data centres could become unusable."
The researchers propose two policy responses. First, they recommend establishing sovereign control over the computing infrastructure needed for critical national functions. Second, they propose working with other technologically advanced middle powers, including Canada, France, Japan and South Korea, to develop the next generation of advanced AI through an international consortium modelled on collaborations such as CERN and Airbus.
Associate Professor in Electrical Engineering, Amro Awad, at Evidence Week 2026
No single middle power is likely to reach the frontier alone. But by pooling investment, talent and infrastructure, countries can collectively build capabilities that would otherwise remain beyond their reach, while strengthening their own research ecosystems.
The briefing formed part of Evidence Week in Parliament, organised by Sense About Science which brings researchers together with MPs and policymakers to help ensure public policy is informed by the best available evidence. For the Oxford Martin School, it was an opportunity to bring Oxford research directly into discussions about one of the defining policy challenges of the AI era.
As AI becomes an increasingly strategic resource, the researchers argue that the question is no longer simply who develops the technology. It is who can rely on using it when it counts.
The team's participation in Evidence Week was supported through Research England's Policy Support Fund via the Oxford Policy Engagement Network (OPEN).