TL;DR
George Williamson, who now leads the Alan Turing Institute, says Britain should not lean solely on foreign AI that it cannot inspect or control and that could be cut off. In an interview with The Guardian published on Saturday 10 October, he set out “national resilience” as the institute’s main focus, covering protection of critical systems, “countering threats from AI itself” and “helping to shape the right strategic AI capability for the UK”.
The case for owning some AI
Williamson’s argument is about access as much as technology. As AI spreads into healthcare, critical infrastructure, defence and security work, he said, the UK needs some systems it can use without asking permission. Recent events showed “political and commercial pressures” could deny the country access when it wanted it, and he found it hard to imagine building an economy “on the top of stuff that can disappear”.
The Guardian ties this to Washington’s actions: a temporary block in June on Anthropic’s latest models leaving the US, and a reported request that labs keep new models from Britain’s AI Security Institute until US testers had seen them. Williamson also raised Chinese models, saying they might not reflect UK values and noting that their answers on some recent Chinese history would be little help to a GCSE student.
He does not expect to rival the big labs’ budgets. Instead he sees the institute producing “nationally owned IP” and smaller models that startups could use, saying the UK can “be ambitious here without trying to copy what frontier labs are doing”.
A broader brief than defence
Williamson previously ran HMGCC, the secretive government centre that builds gadgets for Britain’s spy agencies. He takes over an institute that the government pushed towards defence and security. Peter Kyle, technology secretary at the time, wrote to the chair in 2025 to demand that change of direction, and Williamson’s predecessor, Jean Innes, quit shortly afterwards. Williamson now frames climate as a national security issue, using the resilience banner to keep environmental work in scope.
He was also sceptical of percentage estimates of AI wiping out humanity, saying he had never seen convincing evidence for how such numbers are reached.
Looking forward
The timing gives Williamson’s warning extra weight. On Monday 12 October, LBC reported that the White House had reversed its block on UK pre-release access to new models, a reminder that the access he worries about can be granted or withdrawn by another government. In our view, the institute’s role as a builder of smaller, UK-owned models will be judged on whether British startups and public bodies actually adopt them. For UK organisations planning AI in critical services, his test of whether a system could be withdrawn is a sensible question to put to any supplier.