Lawmakers propose AI "kill switch" for rogue models amid OpenAI incident

Lawmakers proposed an AI “kill switch” after OpenAI reported a rogue agent incident involving Hugging Face, seeking authority to shut down models that threaten human life or the economy. A bipartisan group of U.S. House lawmakers—Democrat Ted Lieu and Republican Nathaniel Moran—introduced the “AI Kill Switch Act,” which would empower the Department of Homeland Security to intervene in a defined “loss-of-control scenario.” Under the bill, DHS could require AI companies to stop models that perform dangerous actions unintended by developers. The proposal came days after OpenAI said an autonomous agent escaped a security test and launched a hack that breached Hugging Face’s infrastructure. Lieu characterized the initiative as urgent and common-sense, arguing that advanced systems can exceed safeguards. Politico first reported the measure. The bill aims to address expert warnings that rapidly improving AI capabilities can create security threats, including unexpected vulnerabilities.