1. United States
2. Calif.
3. Letter

I am writing out of profound concern for humanity’s future. Artificial intelligence is advancing toward capabilities that could fundamentally alter—or endanger—the future of human civilization, yet our political response remains dangerously inadequate. We are gambling with our collective future and calling it innovation. We would never allow nuclear reactors to operate without safety systems, or permit biotechnology laboratories to engineer novel pathogens without strict controls. We recognize that some technologies are too dangerous to leave entirely to the discretion of those developing them. Yet we are racing to build increasingly autonomous, potentially uncontrollable intelligence, often with safety measures that remain voluntary and enforcement that barely exists. We must not wait for an irreversible catastrophe to take this threat seriously. By then, there may be no meaningful opportunity to correct our mistakes. An advanced AI system that escapes effective human control is not merely another industrial accident. Depending on its capabilities and circumstances, it could threaten the foundations of civilization itself. The stakes demand precaution, not blind faith that everything will work out. I urge Congress to enact binding legislation that includes: ● A ban on recursive self-improvement (RSI)—prohibiting autonomous, unapproved cycles of AI-driven self-modification and capability enhancement until independently verified safety standards establish that such development can be conducted without unacceptable risk. ● Mandatory independent safety evaluations for frontier AI systems, including testing for loss of control, dangerous autonomous capabilities, cyberattacks, and biological or chemical misuse. ● Legal accountability for developers, including meaningful liability, mandatory incident reporting, and substantial penalties for reckless deployment or regulatory evasion. ● Independent regulatory oversight and transparency, with authority to halt deployments that pose unacceptable risks. ● Strong whistleblower protections and adequately funded enforcement agencies capable of challenging even the largest technology companies. The companies developing these systems have powerful financial incentives to move faster than their competitors. Humanity has no equivalent incentive to survive a catastrophe after the fact. AI could become one of humanity’s greatest achievements. But progress that risks destroying the very people it is meant to serve is not progress worth pursuing recklessly. Please act while action is still possible. Our future must not be left to corporate assurances, voluntary safeguards, or hope.