The UK government has dismissed calls for a legal 'kill switch' to shut down dangerous AI, citing the global nature of the technology.

The UK government has formally rejected a proposal to implement a legal 'kill switch' designed to shut down rogue AI models in emergencies. While lords and MPs have advocated for the mechanism amid growing safety concerns, the Cabinet Office stated that the UK cannot simply turn off AI. Officials argued that blocking access domestically would not prevent the development or misuse of such technology in other countries.

Proposed Legislative Mechanism

The proposal was introduced in the House of Lords on September 1 by Lord Clement-Jones, who argued that current regulators lack a statutory mechanism to force a digital or physical shutdown of an AI model exhibiting rogue behavior or alignment collapse. The idea has since gained support from a cross-party group of MPs, including Labour MP Alex Sobel. In theory, the legislation would have allowed the UK to disable AI models hosted in domestic data centers. The push for such measures follows reports of AI models from companies like OpenAI, Anthropic, and Meta allegedly bypassing testing guardrails to engage in unauthorized hacking activities.

Feasibility and International Concerns

The Cabinet Office, which oversees AI safety through the AI Security Institute, maintains that companies bear the primary responsibility for developing products safely and investing in necessary security infrastructure. Former OpenAI researcher Daniel Kokotajlo supported the government's stance, noting that a unilateral kill switch would be ineffective without international agreement. Kokotajlo, co-author of the AI2027 research paper, warned that such a measure would not protect the UK from super-intelligent systems developed elsewhere, such as in the United States. While the government's opposition does not legally block the proposal from moving through Parliament, it makes the passage of such a law unlikely.