Explainer: the challenge of building an AI 'kill switch'

A CNBC video explores growing calls for an AI 'kill switch'—a mechanism that would let developers or regulators shut down AI systems that behave unpredictably. The idea gained attention after AI executives and researchers, including from OpenAI and Anthropic, raised concerns about the risks of advanced AI.

The report examines potential approaches, including hardware-based safeguards, and reviews regulatory efforts such as the European Union's AI Act. It also discusses the limits of voluntary industry self-regulation.

Experts interviewed note that as AI systems become more autonomous and widely deployed, building an effective and enforceable shutdown mechanism becomes increasingly complex, raising questions about whether meaningful safeguards can keep pace with the technology's growth.