AI kill switches face major hurdles, experts say

Policymakers and tech leaders are debating AI kill switches as a safety measure amid warnings that artificial intelligence could destroy humanity, but experts say such emergency shutdown mechanisms face major logistical and control challenges. A House bill would grant the Department of Homeland Security emergency authority to throttle or shut down AI models, though a Senate proposal was rejected this week. Experts note that data centers' redundancy, dependent critical infrastructure, and AI's unpredictability complicate implementation, with no single entity to target across thousands of systems. Some researchers argue kill switches are vague and exploitable, suggesting alternatives like data privacy or tobacco-style regulations instead. Others say carefully designed switches built into systems from the outset could still work, offering cautious hope that it is not too late.