The UK Cabinet Office has rejected calls for a single, overarching "kill switch" to be used to shut down dangerous artificial intelligence, saying the country "cannot simply turn AI off". The department, which leads on AI safety across government, framed the proposal as impractical in light of how modern systems are deployed and integrated.
Proposals for an emergency shutdown mechanism have gained attention in recent months as part of a wider debate over how best to manage risks from increasingly capable AI. Proponents argue that a clear technical fail-safe could provide a backstop against runaway or harmful behaviour, while critics caution that a single switch may be technically difficult to implement and could be circumvented or cause unintended disruption.
The Cabinet Office statement underscores the practical challenges policymakers face when translating high-level safety goals into enforceable measures. Rather than relying on a single technical button, the government appears to be signalling a preference for other forms of oversight and risk management as the focus of its safety work.
The question of how to govern advanced AI is under active discussion internationally, with governments, industry and academics exploring a range of regulatory, technical and operational responses. The UK government's rejection of a one-size-fits-all shutdown mechanism is likely to sharpen debate about which combinations of safeguards — regulatory standards, testing, transparency and incident response arrangements — will be most effective as systems become more capable.
As the technology evolves, policymakers will continue to weigh the technical feasibility of particular interventions against the need to manage harms. The Cabinet Office's position makes clear that, for now, the UK is sceptical that a single emergency off-switch is a practical route to ensuring AI safety.