Set Enforceable Safeguards Before Advanced AI Causes Serious Harm

As a constituent, I urge Congress to establish enforceable safeguards for the most powerful artificial-intelligence systems. Systems capable of catastrophic harm should not advance without meaningful human control, independent testing, and public accountability. HUMAN BEINGS MUST REMAIN IN CONTROL AI should serve human beings, not create systems people can no longer understand, stop, or control. The greatest concern is not ordinary AI. It is powerful systems that could affect critical infrastructure, cybersecurity, weapons, biological security, and other areas where failure or misuse could endanger many people. The National Institute of Standards and Technology (NIST) has identified AI risks ranging from cyberattacks and biological misuse to manipulation, privacy harms, and system failure. Its AI Risk Management Framework is voluntary, and NIST is developing guidance for critical infrastructure. NIST’s work provides a foundation Congress can build on when setting enforceable safeguards for the highest-risk systems. COMPETITION DOES NOT REQUIRE RECKLESSNESS The New York Times Opinion article “The ‘But China!’ Dilemma Driving the A.I. Race,” published September 15, 2026, examines whether stronger U.S. safeguards could let China gain an advantage. That concern is real, but competition cannot justify abandoning prudent protections. A race that increases danger to Americans, critical systems, or humanity would not make the United States more secure. The public should not have to rely on voluntary promises by AI companies. Developers are rewarded for moving quickly. When being first brings major commercial and strategic advantages, public safety should not depend entirely on voluntary restraint. Innovation matters, and the United States should remain an AI leader. But leadership should not require gambling with human control over dangerous systems. I urge Congress to enact the safeguards below before a major failure demonstrates why they were necessary. (1) REQUIRE SAFETY TESTING. Require independent pre-deployment testing and continuing evaluation for the most powerful AI systems. (2) ENSURE HUMAN CONTROL. Require developers to show that people can reliably stop, override, or contain high-risk AI systems. (3) REQUIRE INCIDENT REPORTING. Mandate prompt reporting of serious failures, breaches, dangerous capabilities, and loss-of-control events. (4) CREATE ENFORCEABLE OVERSIGHT. Designate a federal agency with authority to set and enforce safety standards for the highest-risk AI systems, subject to congressional oversight. (5) PROTECT CRITICAL DECISIONS. Require meaningful human authorization for AI used in nuclear weapons, lethal military systems, critical infrastructure, and other decisions where autonomous errors could threaten life or national security. (6) PURSUE INTERNATIONAL SAFEGUARDS. Seek verifiable agreements with China and other major AI powers on human control of weapons, catastrophic biological and cyber risks, and emergency notification. Thank you.