AI 'kill switch' may need to be mandatory, Anthropic co-founder says - BBC

Jack Clark, a co-founder of AI firm Anthropic, told the BBC that a mandatory "kill switch" allowing AI systems to be shut down completely if they become too dangerous may need to be enforced by lawmakers.
Clark said most labs, including Anthropic, already have ways of "pulling the plug," but suggested society may want rules requiring such a mechanism and asked whether it should be verifiable by a third party.
Anthropic chief executive Dario Amodei recently called for the pace of AI development to slow and be more closely monitored, though he added any action should come "without sacrificing commercial advantage," prompting some to question the motivations.
Fears over AI risks have intensified after an Anthropic researcher who quit over concerns about human extinction went viral; Anthropic scientist Evan Hubinger said he personally put the chance of AI-driven human extinction at over 10% within the next decade.
Nobel Prize winner Geoffrey Hinton, dubbed the "Godfather of AI," told the BBC a 10% chance of AI killing all humans was "not unreasonable," with some warning AI could seize control of internet-connected systems.
Skeptics argue such extinction fears are exaggerated or intended to generate hype, with Hugging Face leader Clement Delangue dismissing the framing of the debate.