Senator Cantwell outlines a federal framework for frontier AI safety

Senator Maria Cantwell, the Washington Democrat and ranking member of the Senate Commerce, Science and Transportation Committee, outlined a governance framework for frontier AI built around six principles, rather than introducing a bill. She said the aim was to guide “legislative and executive actions needed to address catastrophic risks and other AI harms,” building on positions she has held since chairing the committee in earlier years.

The framework calls for the National Institute of Standards and Technology, working with other federal agencies, to set clear, measurable, risk-based safety standards for systems capable of catastrophic harm — covering AI-enabled cyberattacks, chemical, biological, radiological and nuclear threats, loss of human control, and autonomous agents escaping secure testing environments — and for covered models to undergo an independent audit confirming they meet those standards before release. It calls for continuous government and independent testing of deployed models, with developers required to give qualified outside testers and auditors the access needed to investigate incidents; for mandatory, prompt reporting of serious safety and security incidents, including loss of control over agents; for developers to remain civilly and criminally liable for foreseeable harms; for whistleblower protections; and for public-private partnerships and US-led international cooperation on common safety standards.

To manage risks from advanced AI systems we need clear safety standards, continuous testing, and reporting of serious failures. America can lead the world in AI by building systems that are not only more capable, but safer and more secure.

— Maria Cantwell

As a framework rather than a bill, the announcement sets out principles Cantwell wants reflected in future legislation and agency action without proposing specific statutory text, leaving its practical effect dependent on whether Congress or federal agencies take it up. It arrives amid continued argument in Washington over whether frontier AI safety should be governed by binding federal standards, administration-led voluntary measures, or left to the states.