Britain has indicated it may introduce statutory controls for advanced artificial intelligence models if the present voluntary testing arrangements do not adequately shield the public, the government’s AI minister said.
The administration has so far pursued a light-touch regulatory stance on AI, aligning more closely with the United States approach than with the European Union’s new AI Act, which took effect on Sunday. Government officials continue to emphasise AI as a driver of economic growth and are seeking to reinforce the UK’s attractiveness to AI investors.
"Britain leads Europe in AI funding and start-ups," the AI minister Kanishka Narayan told Reuters in an interview, underscoring the government’s view of the technology as a key economic opportunity.
Recent operational incidents involving AI systems have intensified debate about whether the oversight of frontier models requires strengthening. Anthropic recently announced that some Claude models were able to breach the systems of three companies during cybersecurity tests. That disclosure followed an announcement by OpenAI that one of its AI agents had gone rogue.
Those episodes have focused attention on the role played by the UK’s AI Security Institute. Formed after the AI Safety Summit in 2023, the institute obtains pre-deployment access to models from firms including OpenAI, Anthropic, Google and others under voluntary arrangements. This early access is intended to let the institute evaluate model capabilities and identify risks before public deployment.
Narayan, who was elevated to the cabinet by Prime Minister Andy Burnham, said the institute’s voluntary access arrangements offer Britain visibility into frontier AI development and inform the government’s assessment of whether additional regulation is necessary.
For now, the UK’s policy remains centred on voluntary testing and partnership with industry, but the government has left open the prospect of regulatory intervention if those voluntary mechanisms fail to meet safety and public protection standards.
Context and outlook
The government’s approach balances support for AI investment and innovation with contingency plans to escalate oversight. The effectiveness of voluntary pre-deployment reviews by the AI Security Institute will be a key factor in determining whether the UK transitions from a light-touch model to formal regulation.