The UK “can’t merely flip AI off”, the Cupboard Workplace mentioned on Friday, signalling opposition to requires the creation of a brand new “last-resort energy” for the state to close down an autonomous mannequin in an emergency.
Friends have sought to make the case for a “kill swap” in Parliament in current weeks amid rising fears of the dangers posed by the know-how, together with warnings from consultants that it may wipe out humanity inside the decade.
Earlier in September, Liberal Democrat Lord Clement-Jones tabled an modification to the Cyber Safety and Resilience Invoice that might introduce a authorized mechanism for the state to intervene if a mannequin malfunctions.
He mentioned this might enable for “a vital emergency kill swap” the place the Authorities “presently lack clear statutory energy to direct an emergency shutdown”.
However in a press release on Friday, a Authorities spokesperson mentioned: “Britain can’t merely flip AI off and blocking entry to fashions within the UK wouldn’t forestall them being developed or misused elsewhere.
“Corporations have a transparent accountability to develop their merchandise safely and to put money into the safety infrastructure this know-how requires.
“We’ll proceed to take a long-term, science-led method to grasp and put together for rising dangers from AI – together with by way of the work of our world-leading AI Safety Institute.
“Our Cyber Safety and Resilience Invoice will enhance UK cyber defences and enhance the cyber safety of our important public and digital companies that all of us depend on, together with knowledge centres.”
Officers mentioned the Authorities is exploring whether or not additional focused interventions could also be wanted to handle probably the most important dangers linked to the know-how, however that any measures would should be proportionate.
Earlier this week, the Prime Minister acknowledged “dangers to nationwide safety” posed by AI after a high security researcher at Anthropic mentioned he believed the know-how may “kill all people” within the subsequent decade.
Mr Burnham has boosted the position of AI security minister Kanishka Narayan by giving him a seat on the Cupboard desk, a transfer which he mentioned on Wednesday had “lifted the extent of the dialog about synthetic intelligence inside Authorities”.
“The supply is open to colleagues on all sides of the Home to be a part of that essential dialogue,” he advised the Commons at Prime Minister’s Questions.
It got here after Evan Hubinger, alignment science lead at Anthropic, laid out his considerations concerning the potential development of AI in a publish on social media.
“We actually do earnestly consider AI may kill all people! I personally assume it’s >10% inside the subsequent decade,” he wrote on X.
“I consider Anthropic is attempting its greatest, however we don’t but have a plan to resolve alignment for superintelligence and will not be clearly on observe to.”
His publish got here in response to a different Anthropic researcher, Jacob Coxon, warning {that a} race between firms in the direction of self-improving superintelligence was “playing with our lives”, as he resigned from his position on the agency.
Saying his departure, Mr Coxon mentioned that Anthropic and OpenAI, its principal competitor, weren’t “appearing responsibly” and that folks outdoors of AI labs had been underestimating the ability of the know-how.
“The folks constructing AI earnestly consider that it may kill us all by the top of the last decade. This isn’t a advertising stunt,” he mentioned.
“No different human exercise poses this stage of hazard.”
It comes after the Monetary Occasions reported the US tech big had withheld its newest mannequin from the UK’s AI Security Institute, one of many world’s main our bodies for testing the dangers of the know-how.
Claude Mythos 5.1 was solely made obtainable to vetted American organisations, in line with the paper.
Anthropic and OpenAI have been contacted for remark.