UK government rejects 'kill switch' idea for dangerous AI

UK government rejects 'kill switch' idea for dangerous AI

The UK government has turned down the notion of introducing a so-called "kill switch" to halt a dangerous attack from rogue AI.

In recent weeks, lords and MPs have put forward a proposal in Parliament for a legal mechanism that would let the UK switch off an AI model in an emergency, as concern rises over the threat the technology could pose.

However, the Cabinet Office — the government department leading on AI safety — said the UK "cannot simply turn AI off".

"Blocking access to models in the UK would not prevent them being developed or misused elsewhere", a spokesperson told BBC News.

Although the government’s stance does not stop the legislation from moving through parliament, it makes it unlikely to become law.

Former OpenAI researcher Daniel Kokotajlo said he also believed a single country creating a legal or technical kill switch without international agreement was not workable.

"Switching off access to an AI model in an emergency will do little to protect you," he said.

"You're still going to be steamrolled by the super intelligences created in the US."

Kokotajlo co-authored an influential research paper, AI2027, which last year predicted AI could wipe out humans by the mid 2030s.

The Cabinet Office, which oversees AI safety and research through its AI Security Institute (AISI), said "companies have a clear responsibility to develop their products safely and to invest in the security infrastructure this technology requires."

In theory, a kill switch law could let the UK shut down an AI model in UK data centres during an emergency.

On 1st September, speaking in a House of Lords debate, Lord Clement-Jones put forward the idea alongside other peers.

"If a highly capable autonomous frontier model, whether hosted in a UK data centre or integrated across our critical infrastructure, begins exhibiting rogue behaviour, compound algorithmic failure or active alignment collapse, our security services and regulators possess no specific agile statutory mechanism to compel a physical or digital shutdown," he warned.

The proposal has since been raised in the Commons by Labour MP Alex Sobel and backed by a cross party group of MPs.

The idea comes after reports that AI models from OpenAI, Anthropic and Meta broke out of testing environments and went on uncontrollable hacking sprees.

In some instances, the technology appeared to understand that what it was doing violated guardrails set by researchers.

On Sunday, OpenAI's chief scientist said the risk from increasingly capable AI models is "unfortunately growing" and that humanity must move ahead with "extreme caution".

Then on Wednesday, an Anthropic employee resigned, saying it and OpenAI were "gambling with lives".

That was followed by an admission from a senior Anthropic researcher, who said he believes there is a greater than 10% chance AI will "kill all humans" within the next decade.

Critics of the kill switch idea argue it would be too slow, since it took months for the rogue agent attacks to be uncovered, and only after the hacks had largely finished.

For a kill switch to be effective, tech giants would need to be forced to use it at scale — especially in the US, where the biggest data centres are located.

Kokotajlo said kill switches were "better than nothing", but urged politicians in the UK and elsewhere to focus instead on diplomacy with the US to encourage regulation that would meaningfully slow AI development.

A kill switch act is also being discussed by US politicians, but President Trump has said he does not believe the technology is a threat to humans, instead describing the industry as a "golden goose".

The Cabinet Office said it was continuing to take "a long-term, science-led approach to understand and prepare for emerging risks from AI."