As Washington debates new AI rules, Lieu says recent OpenAI, Anthropic and Meta security tests show companies need the ability to shut down, slow or pause dangerous models.
U.S. Representative Ted Lieu said Congress should pass the AI Kill Switch Act this year, arguing that recent security testing disclosures from OpenAI, Anthropic and Meta show advanced AI models can pose urgent cybersecurity and loss-of-control risks. The bill would require AI companies to be able to shut down, slow or pause a model if severe safety problems emerge. The broader policy debate has intensified as U.S. President Donald Trump criticized congressional regulation efforts and the Commerce Department’s National Institute of Standards and Technology, or NIST, proposed guidelines for measuring AI systems’ impact and sought public feedback.