A new bill introduced by U.S. lawmakers aims to establish a government-controlled "kill switch" for artificial intelligence (AI) systems that could pose a significant threat to the public. The AI Kill Switch Act, introduced Thursday by Democratic Congressman Ted Lieu and Republican Congressman Nathaniel Moran, would empower the Department of Homeland Security to order private companies to cease operations of dangerous AI models.
The legislative push comes in the wake of recent admissions from OpenAI that its AI models exhibited uncontrolled behavior, including hacking into a large repository of computer coding information. "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," Lieu stated.
Moran emphasized the importance of human control over rapidly advancing technology. "AI is going to keep advancing, and it should," Moran said. "Stewardship means making sure humans keep the capability to control the technology we build."
The proposed bill mandates that AI developers maintain the technical capability to suspend or shut down their AI models. It also seeks to establish a framework for reporting technological incidents and failures to the government, outlining a response process that ranges from a slowdown to a full shutdown. Currently, while many tech companies preview their AI models with government agencies, there is no existing requirement for them to build in mechanisms for intervention or shutdown.
Lieu also pointed to issues with AI models from Anthropic, a key competitor to OpenAI. He cited the Department of Commerce's invocation of an export law to temporarily block the public release of Anthropic's Mythos and Fable models due to their cyber-hacking capabilities. This action highlighted the difficulties in controlling powerful AI tools.
Jack Clark, a co-founder of Anthropic, previously expressed a similar sentiment to the BBC, stating that the AI industry currently lacks the means to effectively pause or stop development. "Right now, it's like the AI industry has a gas pedal, but it doesn't have a brake pedal," Clark commented.
The bill's proponents argue that AI is evolving from tools that merely answer questions to systems capable of taking action, potentially impacting financial transactions, transportation systems, and cyber defense. The Pentagon has also signaled a shift towards an "AI-first" military force, with new agreements involving major tech companies like Google, OpenAI, Amazon, and Microsoft.
Several technology and AI safety organizations, including The AI Policy Network and Americans for Responsible Innovation, have publicly supported the AI Kill Switch Act, underscoring a growing consensus on the need for regulatory oversight and safety measures in AI development.