UK science advisers urge domestic AI investment to avoid dependence on foreign suppliers

Britain needs urgent action to strengthen its own AI capabilities or risks becoming a "high-value AI customer" with little influence over the technology, the government's top science advisers have warned.

Kevin Osei
lake-dbc4f8c8-3b41-478c-b780-d8048d614aa1-0.jpg

A call for greater control

The Council for Science and Technology made the warning in a letter to AI Minister Kanishka Narayan dated July 30, 2026, and published by the UK government on October 19. The independent advisory body's co-chairs, Dame Angela McLean and Lord Browne of Madingley, signed the letter.

The council said Britain's choices now would shape its ability to benefit economically and publicly from AI while preserving resilience and strategic autonomy. Its warning comes as governments compete to develop the expertise, computing resources and models needed to limit dependence on a small group of dominant, mostly US-based companies.

"The way for us to achieve AI sovereignty is to build trusted systems that deliver real public, economic and security benefits," the advisers wrote. They called for "deliberate and urgent action" to prevent Britain from falling behind.

The letter was also circulated to the prime minister, the chancellor of the Exchequer, the secretaries of state responsible for business, defense, and digital, culture, media and sport, the chair of the Prime Minister's AI Taskforce, the cabinet secretary and relevant departmental permanent secretaries, among others. The government did not immediately respond to a request for comment.

A national plan and practical tests

The council put a UK AI Sovereignty Plan first among its recommendations. It proposed beginning with a rapid assessment of computing power, data, skills, models, governance and public adoption. That work would establish what is already happening, expose dependencies and gaps, and guide decisions about which capabilities Britain should own, develop with partners or obtain through resilient access arrangements.

To coordinate the effort, the advisers proposed one accountable policy lead, potentially the head of the Cabinet Office AI Taskforce. The role would bring departments together and help guide funding and delivery decisions.

They also recommended testing AI systems in national security, health and education. Each public-interest pilot should define the problem it seeks to solve, identify the sovereign data Britain holds, specify the model capabilities and surrounding system required, establish what value could be captured, and set measures of success.

Longer-term investment would include an AI skills and secondment route alongside the government's TechFirst program, plus continued backing for fundamental computing research, including low-power approaches. The council also proposed trials of data access arrangements in at least two high-value areas, with health and creative data as examples.

Domestic capabilities alongside global access

The advisers did not call for Britain to give up access to leading international AI systems. They recommended retaining that access while building selected capabilities at home, including considering a publicly governed British model.

One option is the Alan Turing Institute's proposal for an open-source sovereign base model released as a public good. The council urged the government to consider the proposal, which treats the model as national infrastructure under UK governance rather than a proprietary product belonging to one company or department. Researchers, startups, small and medium-sized businesses, public services and the open-source community could use it.

Existing funding and model development

Britain had made sovereign AI funding commitments before the council's advice. The Sovereign AI Unit has backing of up to £500 million; its next phase launched in April 2026, chaired by James Wise of Balderton Capital.

The government also announced a £1.1 billion AI Hardware Plan at London Tech Week 2026. At its center is a £750 million national AI supercomputer scheduled to go live in 2030.

Cosine, a company selected under the £500 million Sovereign AI program, is training the frontier coding model Lumen Sovereign with a coalition of UK institutions. Participants include BT, Lloyds, NatWest, LSEG, PwC, BAE Systems, Leonardo UK, Babcock and the Alan Turing Institute. Delivery is targeted for late 2026.

← Back to Artificial Intelligence

Related news