The UK government has rejected parliamentary proposals to establish an emergency "kill switch" for rogue artificial intelligence, stating that the state cannot simply turn AI off and that domestic blocks would not stop development or misuse abroad. The decision comes despite warnings from researchers at firms like Anthropic and OpenAI that superintelligence poses existential risks to humanity. Critics and experts agree that unilateral national shutdown mechanisms are unfeasible without international coordination.
UK kill switch proposal rejection
- ▪The UK Cabinet Office stated that the UK cannot simply turn artificial intelligence off, arguing that blocking domestic access would not prevent models from being developed or misused in other countries.
- ▪The UK Cabinet Office rejected a proposal on September 11, 2026, to establish a legal "kill switch" mechanism that would allow the state to shut down autonomous artificial intelligence models during an emergency.
Parliamentary emergency shutdown advocacy
- ▪Labour MP Alex Sobel and a cross-party group of MPs supported the emergency AI shutdown proposal in the House of Commons.
- ▪Liberal Democrat peer Lord Clement-Jones submitted an amendment to the Cyber Security and Resilience Bill in early September 2026 to create a legal mechanism for state-directed emergency shutdowns of malfunctioning AI models.
AI existential risk warnings
- ▪The AI2027 research paper, co-authored by former OpenAI researcher Daniel Kokotajlo, predicted that artificial intelligence could wipe out humanity by the mid-2030s.
- ▪Anthropic alignment science lead Evan Hubinger stated on social media that he believes there is a greater than 10% chance that artificial intelligence could kill all humans within the next decade.
- ▪Former Anthropic researcher Jacob Coxon resigned from the firm, warning that a race between companies toward self-improving superintelligence is gambling with human lives.
International feasibility concerns
- ▪Former OpenAI researcher Daniel Kokotajlo stated that a single country creating an AI kill switch without international agreement is unfeasible and would do little to protect against superintelligences developed in the United States.
- ▪Critics of the AI kill switch proposal argue that such a mechanism would operate too slowly, noting that previous rogue agent attacks took months to discover and were only identified after the hacks had largely concluded.
UK AI safety governance
- ▪The Financial Times reported that US technology firm Anthropic withheld its Claude Mythos 5.1 model from the UK's AI Safety Institute, making it available only to vetted American organizations.
- ▪UK Prime Minister Andy Burnham elevated the role of AI safety minister Kanishka Narayan by granting him a seat at the Cabinet table to raise the level of discussion about artificial intelligence within the government.
Debatable claims
- ▪The UK should establish a legal emergency 'kill switch' for autonomous AI models
- ▪AI safety efforts should focus on international diplomacy rather than unilateral national controls
Story comments
Loading comments…