George Williamson, head of the Alan Turing Institute, warned in an interview published on 10 October that the UK should develop its own AI capability rather than depend entirely on foreign systems that could be restricted or switched off, The Guardian reported. He said national resilience would become a focus for the institute, which is the UK’s leading AI research body and receives most of its funding from the government.
Key points
- Williamson wants the UK to retain access to AI systems it can inspect and control.
- Washington temporarily blocked the release of Anthropic’s latest models outside the US in June.
- He says smaller, application-specific models could give the UK useful capability at lower cost.
- The institute’s proposed resilience work spans critical systems, AI threats and public uses of the technology.
Washington’s restrictions sharpen Williamson’s warning
Williamson identified access as a practical concern for services that increasingly use AI. “As AI becomes more embedded in critical services, in domains like defence, national security, critical infrastructure and healthcare we shouldn’t depend solely on systems we can’t inspect or control, or that could be restricted or even switched off,” he said. He argued that political and commercial pressures could prevent the UK from using AI when it needed it.
Washington temporarily prevented Anthropic’s latest models from being released outside the US in June, The Guardian reported. It has also reportedly asked leading AI laboratories to keep their models from testing by the UK’s AI Security Institute until its US counterpart has examined them. The first restriction concerned access to models, while the reported request concerns the order in which national testing bodies can examine them.
Williamson said the UK needed more control over its access to technology and a stronger domestic position in AI. He also cited concerns that Chinese-made systems might not align with UK values, giving the example of a student seeking answers about events in China’s recent history. Christine Lagarde, president of the European Central Bank, and Ursula von der Leyen, president of the European Commission, have urged Europe to become more self-sufficient in AI. Lagarde spoke of the risk of being cut off from US and Chinese technology.
Kyle’s 2025 letter and the Turing remit
The institute’s change in direction began before Williamson’s warning. In 2025, Peter Kyle, then the UK’s technology secretary, wrote to its chair demanding that defence and security become its main focus, The Guardian reported. Jean Innes, Williamson’s predecessor, resigned soon afterwards. Williamson previously headed His Majesty’s Government Communications Centre, a government body that makes devices for UK intelligence agencies.
Williamson described a broader remit for national resilience. He said the institute would work on protecting critical systems against disruption, countering threats arising from AI and helping determine what strategic AI capability the UK needs. He also proposed work on weather forecasting, farming and the detection of disinformation. Andy Burnham, the UK prime minister, announced a new national centre to tackle state-backed disinformation in September.
Climate research could remain within that remit. Williamson called climate as much a national security issue as defence and said resilience would allow the institute to address a range of threats and opportunities. He also described the institute, a registered charity, as a potential arbiter in public debates about AI risks. On estimates of AI-triggered extinction, he questioned the evidence behind the percentages used by some in the sector.
Smaller models and the Turing board
Williamson said the UK and the institute probably could not match the financial resources of major AI laboratories. He said the UK and the institute could produce nationally owned intellectual property or smaller models for startups, rather than try to copy frontier laboratories. He said the institute’s work suggested application-specific models could come very close to frontier performance with substantially lower computational and financial demands, Quantum Zeitgeist reported.
Williamson also warned that some highly capable AI technologies could be extraordinarily expensive to deploy at scale in public services. The question of how to build domestic capacity extends beyond the institute: AI Affairs previously reported that the UK’s £500m Sovereign AI Fund met Northern Ireland founders in Belfast.
The institute appointed Tracey Bowen-Lowe, Catherine Graham and Rakshit Kapoor as non-executive directors and trustees, EdTech Innovation Hub reported on 7 October. Bowen-Lowe has worked in technology and consulting, Graham is a former chief financial officer of Darktrace, and Kapoor is Chubb Insurance’s global chief data officer. Graham said she would help steer the institute’s work on resilience and security.
