Jack Clark, a co-founder of Anthropic, has told the BBC that AI companies may need to be legally required to install a verifiable "kill switch" allowing dangerous AI systems to be shut down completely. He said this was the kind of measure society might eventually need to legislate for, as policymakers grapple with growing concern from AI executives, staff and researchers about the technology's risks, including warnings that AI could ultimately threaten humanity's survival.

Clark said most AI labs, including Anthropic, already have some means of "pulling the plug", but suggested regulation could make this mandatory and independently verifiable. His comments follow Anthropic chief executive Dario Amodei's call for AI development to slow down, and remarks from Anthropic scientist Evan Hubinger, who put the chance of human extinction from AI at over 10% within a decade — a figure Nobel laureate Geoffrey Hinton also called "not unreasonable". In the US, a proposed "Kill Switch Act" would compel firms to build in shutdown capabilities, though President Donald Trump has dismissed AI-safety fears as a "hoax", while the UK government has also rejected mandating a kill switch, saying it would not stop AI being developed or misused elsewhere.