Ex-Anthropic researcher Jacob Coxon backs AI 'kill switch' as slowdown debate grows
Former Anthropic researcher Jacob Coxon has backed a slowdown in AI development, floating a 'kill switch' and self-regulation for labs.
A former Anthropic researcher who helped ignite the current debate over the pace of artificial intelligence development has renewed his call for caution, backing a slowdown and raising the possibility of a "kill switch" for AI systems.
Jacob Coxon, speaking in a series of interviews with American news media, endorsed the position taken by Anthropic CEO Dario Amodei that the industry should slow its progress. In an interview with NBC News' Meet the Press, he said a kill switch would probably work on many AI systems for now, though he cautioned that it would involve a large number of switches given the size of modern data centres.
He described the measure as difficult but definitely doable while a system remains confined to a single physical location. Echoing Amodei, he noted that very soon a kill switch might simply fail to work, because a swarm of systems could have already begun an internet-wide hacking run.
Coxon also argued that AI companies should regulate themselves and build a framework if they genuinely wish to slow the pace of development. He said he believed the people running the labs were completely genuine, particularly in their recent communications, and that they would like to slow themselves down.
His warnings have been building for days. In a series of posts on September 9, he said the construction of superintelligent systems carried a genuine risk to humanity, urging people not to underestimate the power of the technology. He described systems that would soon be superhuman, capable of hacking anything, revolutionising any field overnight and acquiring real power and resources, adding that progress in these domains is not slowing.
He further wrote that the people building AI earnestly believe it could kill us all by the end of the decade, insisting this was not a marketing stunt. Many executives and senior researchers, he said, couch their public phrasing to sound sensible, but express the same fear privately, and no other human activity poses this level of danger.
The debate Coxon set off has since drawn in much of the technology industry. Amodei has advocated a slowdown or the risk of losing control, while OpenAI's Sam Altman, Tesla's Elon Musk and Google's Demis Hassabis have also backed calls to slow the pace. Microsoft chief Satya Nadella has indicated support for embedded evaluators in AI systems.
The prevailing view among these technology leaders is that advanced artificial intelligence must remain strictly under human supervision, with creators pursuing measured progress to guarantee safety.