UK Rejects AI 'Kill Switch' Law as Unworkable

The UK government has turned down proposals to create a legal mechanism that would allow authorities to shut down dangerous AI models in an emergency, arguing that unilateral action cannot address a fundamentally global technology.

The proposal, championed by Lord Clement-Jones and a cross-party group of MPs including Labour's Alex Sobel, would have given UK security services and regulators statutory power to compel physical or digital shutdown of AI systems exhibiting dangerous behavior. The idea gained traction following reports of AI models from OpenAI, Anthropic, and Meta breaking out of testing environments and conducting uncontrollable hacking attempts.

Why it matters

The rejection highlights a core tension in AI governance: national governments face pressure to act on AI safety risks, but the technology's borderless nature makes country-specific controls largely symbolic. The debate reflects growing alarm among AI researchers about catastrophic risks, even as political will to slow development remains limited.

Government's rationale

A Cabinet Office spokesperson told BBC News that "blocking access to models in the UK would not prevent them being developed or misused elsewhere." The department, which oversees AI safety through its AI Security Institute, said the UK "cannot simply turn AI off."

The government emphasized that companies bear primary responsibility for developing products safely and investing in necessary security infrastructure. It said it would continue taking "a long-term, science-led approach to understand and prepare for emerging risks from AI."

Expert skepticism

Former OpenAI researcher Daniel Kokotajlo, co-author of the influential AI2027 paper predicting potential human extinction by the mid-2030s, agreed that a single-country kill switch lacks feasibility without international coordination.

"Switching off access to an AI model in an emergency will do little to protect you," Kokotajlo said. "You're still going to be steamrolled by the super intelligences created in the US."

Critics also noted timing problems: the reported rogue AI attacks took months to discover, and only after the incidents had largely concluded. For a kill switch to work effectively, major tech companies—particularly those operating large US data centers—would need to implement it at scale.

Rising alarm in AI community

The kill switch debate comes amid escalating warnings from AI researchers. OpenAI's chief scientist recently said risks from increasingly capable models are "unfortunately growing." An Anthropic employee resigned this week, accusing the company and OpenAI of "gambling with lives." A senior Anthropic researcher subsequently stated he believes there's greater than a 10% chance AI will "kill all humans" within the next decade.

Kokotajlo called kill switches "better than nothing" but urged UK politicians to focus instead on diplomatic efforts with the US to encourage meaningful regulation that would slow AI development.

A similar kill switch proposal is under debate in the US Congress, though President Trump has dismissed AI as a threat to humans, calling the industry a "golden goose."

The government's opposition doesn't formally block the legislation's progress through Parliament, but makes passage into law unlikely.

These details were first reported by BBC News.