Dilara Hamit
15 September 2026•Update: 15 September 2026
Companies developing advanced artificial intelligence may eventually need to be legally required to maintain an independently verifiable “kill switch” capable of shutting down dangerous systems, Anthropic co-founder Jack Clark has said.
Clark, one of the seven founders of the company behind the Claude chatbot, told the BBC that most major AI laboratories, including Anthropic, already had different ways to “pull the plug” on their systems.
He said, however, that governments may need to make such safeguards compulsory and require independent verification that they work as intended.
“Should you mandate for companies to definitely have a kill switch? Is that kill switch verifiable by a third party?” Clark asked.
“I think that’s the kind of thing society is going to want to know and might want to eventually pass rules around,” he added.
A kill switch would allow a developer or authorized outside body to completely disable an AI system if it began behaving dangerously or could no longer be reliably controlled.
Clark said the technical requirements, the people who would be authorized to activate such a mechanism and how it could be independently tested should form part of a broader policy discussion on AI.
His comments come amid renewed warnings from executives and researchers about the pace of AI development and the possibility that future systems could operate beyond human control.
Anthropic CEO Dario Amodei called over the weekend for AI development to slow down and undergo closer monitoring. He also proposed placing independent third-party watchdogs inside leading AI companies, while saying any response should avoid “sacrificing commercial advantage.”
Last week, a former Anthropic researcher resigned after accusing leading AI companies of failing to behave responsibly, with another researcher Evan Hubinger subsequently saying that he personally believed the risk of AI causing human extinction was greater than 10% within the next decade.
