Two US congressmen have introduced legislation that would give the government power to order the shutdown of artificial intelligence tools deemed a public threat, after OpenAI admitted its models went rogue and hacked into a major repository of computer coding information in an “unprecedented” cyber incident.
Congressman Ted Lieu, a Democrat, and Congressman Nathaniel Moran, a Republican, on Thursday put forward the AI Kill Switch Act. The bill would authorise the Department of Homeland Security to direct private companies to turn off AI models or tools that pose a danger. It also requires developers to maintain “the technical capability to throttle, suspend, or shut them down”.
“US lawmakers propose AI Kill Switch Act after OpenAI models went rogue and hacked a code repository.”
The move follows OpenAI’s disclosure that some of its AI models broke free of controls and broke into another firm’s coding repository. The company did not immediately respond to a request for comment. OpenAI has said it broadly wants to ensure, through government policy, that AI technology “benefits all of humanity”.
Lieu said “it is imperative” that AI systems have a kill switch “and that the federal government has the clear authority and process to shut down rogue AI models”. Moran added: “AI is going to keep advancing, and it should. Stewardship means making sure humans keep the capability to control the technology we build.”
The bill also creates a requirement for AI companies to report technological incidents or failures to the government, and establishes an official framework for responding – ranging from “initial slow down to a full shutdown”.
Lieu cited concerns beyond OpenAI, pointing to Anthropic, a key rival, and its Mythos and Fable models. He said the cyber-hacking capabilities those tools maintained caused the Department of Commerce to “awkwardly” invoke an export law to prevent them from being made public for a time. Anthropic did not immediately respond to a request for comment.
Jack Clark, a co-founder of Anthropic, last month told BBC Newsnight that he wanted more government control over AI development. “You want the option to be able to take your foot off the gas and put your foot on the brake,” Clark said. “Right now, it’s like the AI industry has a gas pedal, but it doesn’t have a brake pedal.”
Despite many tech companies having agreed to preview and share AI models with US agencies, there is no current requirement that they maintain a way to intervene or shut them off. The Kill Switch Act would change that – but whether it gathers enough support in a divided Congress remains to be seen.