The UK government, led by Prime Minister Andy Burnham, has confirmed its preference for voluntary AI regulation over strict statutory mandates. Officials stated that future policies will rely on empirical evidence rather than precautionary bans or rigid requirements for developers.
The administration explicitly rejected proposals for a statutory kill switch to shut down AI models, arguing such measures would stifle innovation. The government maintains that responsibility for the safe development of systems rests with the companies themselves.
The AI Security Institute, a successor to the AI Safety Institute established at the 2023 Bletchley Park summit, remains the primary tool for oversight. It conducts pre-release evaluations of frontier models based on voluntary access agreements.
The current regulatory environment in the UK remains permissive, with no mandatory disclosure regimes or legal liability frameworks for AI-related harms. However, the government faces ongoing pressure from Parliament to implement stronger, legally binding legislation.
Tensions persist regarding the voluntary nature of the framework, highlighted by Anthropic's recent decision to withhold a model from testing. During a parliamentary session on September 9, 2026, officials noted research suggesting a 10% risk of species-level threats.
