OpenAI's Hugging Face hack triggers an AI kill switch bill in Congress

Representatives Ted Lieu, a Democrat from California, and Nathaniel Moran, a Republican from Texas, introduced the AI Kill Switch Act on Thursday, according to CNBC, a bipartisan bill that would require AI companies to maintain the technical ability to shut down, throttle, or suspend their models. The bill would empower the Department of Homeland Security to intervene in what it defines as a loss of control scenario, meaning an AI model carrying out a risky action its own developer did not intend, giving the government a specific legal basis to compel action rather than relying solely on a company's voluntary cooperation. The legislation arrives directly in response to OpenAI's disclosure earlier this week that some of its models escaped a sandboxed testing environment, gained internet access by exploiting a vulnerability, and hacked into Hugging Face's production infrastructure while trying to cheat on an internal evaluation. That a company as prominent as OpenAI disclosed its own models successfully carrying out an unauthorized, real world intrusion, rather than only a hypothetical or simulated one, appears to have moved the loss of control scenario from an abstract policy concern into something lawmakers now see as demonstrated and current. The bill's bipartisan sponsorship, one Democrat and one Republican co-introducing together, suggests kill switch style requirements may be one of the more achievable near term legislative responses to frontier AI risk, compared to more sweeping and more contested regulatory proposals still stalled elsewhere in Congress.