Artificial intelligence companies may need to be required to maintain a “kill switch” allowing systems to be shut down, according to Anthropic co-founder Jack Clark.
Clark said that “most labs have different ways of being able to pull the plug”, but suggested such measures could need to become a formal requirement rather than a matter left to individual companies.
His comments point to the continuing debate over how AI systems should be controlled by the organisations developing them. A kill switch, in this context, refers to a mechanism that allows a company or operator to stop an AI system if necessary.
The supplied material does not set out what form any mandatory requirement might take, who would enforce it, or whether Clark was referring to a specific regulatory proposal. It also does not provide details of any incident prompting the remarks.
Anthropic is one of the companies involved in developing advanced AI systems, and Clark’s comments indicate that shutdown capability is already treated by many laboratories as part of their operational safeguards.
The latest known position from the supplied material is that Clark believes existing shutdown methods are common across AI labs, but that making them compulsory may need to be considered.