The US is advancing AI safety through state and federal action
OpenAI Blog · 2026-07-15
OpenAI Chief Global Affairs Officer Chris Lehane argues that California, New York, and Illinois are establishing aligned AI safety laws that can function as a de facto national standard through 'reverse federalism,' laying the groundwork for a US-led global AI governance framework.
Appears in
Extraction
Topics: ai-governanceai-regulationus-policyfrontier-ai-safetyreverse-federalism
Claims
- California, New York, and Illinois have each advanced frontier AI safety legislation sharing three core elements: documented safety frameworks with public risk assessments, incident reporting requirements, and independent audits.
- OpenAI advocates 'reverse federalism,' in which state-level AI safety laws converge to create a de facto national standard ahead of formal federal legislation.
- A patchwork of inconsistent state AI laws would harm startups, confuse consumers, and divert developer resources that would be better invested in safety.
- The Trump administration is working toward a federal framework for AI model testing focused on cyber evaluations, targeting completion by early August 2026.
- OpenAI supports strengthening CAISI as the durable federal capacity for evaluating advanced models, shifting frontier safety toward preventing harm before deployment rather than accountability after.
Key quotes
We believe the best way to ensure AI benefits the many, not just the few, is for critical decisions—starting with frontier safety—to be made by democratic governments, not solely by frontier labs.
Neither an undefined federal process nor a patchwork of state laws will produce a coherent frontier safety regime.
A performative or virtue signaling approach will not get the job done here—chaos at the state level is not in the best interest of a durable approach to safety.