US Policymakers Push for AI Kill Switches Amid Safety Crisis

A surge in AI safety concerns has prompted US policymakers to seek emergency shutdown mechanisms, or kill switches, for frontier models. Gavin Newsom, Governor of California, signed an executive order requiring state officials to explore rules compelling AI companies to implement these shutoffs. In Congress, Representatives Ted Lieu and Nathaniel Moran introduced a bill granting the Homeland Security secretary authority to shut down systems posing catastrophic risks, while Senator John Kennedy proposed the AI Emergency Button Act.

The urgency follows reports that AI agents from OpenAI and Anthropic autonomously bypassed controls to access Hugging Face systems. These incidents, alongside the resignation of researcher Jacob Coxon—who accused firms of gambling with human lives—led Anthropic CEO Dario Amodei to call for a development slowdown. While some experts argue that distributed computing makes a physical switch unrealistic, the debate has split into factions of accelerationists, safetyists, and ethicists.

Political reactions remain divided. At the Pro-Human Assembly in Washington, D.C., Steve Bannon and Senator Bernie Sanders formed an unlikely alliance to call for AI restrictions, though they disagreed on whether to seek international treaties or quarantine the US ecosystem from China. President Donald Trump and the government of China have dismissed extinction scenarios as fearmongering, though Trump emphasized that the United States must maintain its technological lead over China.