Californians frequently tell themselves that wherever California leads, other states will follow. That was the message echoed by OpenAI in its recent letter to California Governor Gavin Newsom.
The AI giant recently signed the European Union AI Code of Practice, and it’s now asking California state lawmakers to consider such commitments to national and international frameworks compliant with any future legislation. In other words, complying with the EU Code of Practice would mean that companies do not also have to apply with other state regulations that are trying to achieve similar goals. If many other states put similar provisions in their own state AI legislation, innovators would be spared the immense patchwork compliance costs incurred by following a suite of similar-but-not-the-same AI laws.
Consistency in governance is the right goal, but OpenAI is approaching the wrong legislators. State cooperation and its messy outcomes are no substitute for congressional action because states will not all agree to include this one deferential provision.
When Congressional action looked unlikely in 2024, 45 states participated in the Multistate AI Policymaker Working Group (MAP-WG) to craft a cohesive national regulatory framework. The American Consumer Institute (ACI) analyzed 24 algorithmic fairness bills using standard text analysis methods and found drastic differences within the states’ offerings.
At best, these bills were 60 percent similar; at worst, 12 percent. ACI found that states could not agree on fundamental definitions, reporting requirements, and enforcement methods. Only 14 bills affirmed that compliance with the National Institute of Standards and Technology Artificial Intelligence Risk Management Framework (RMF) could constitute a defense in court if a developer were found in violation of the state law. That’s much weaker than OpenAI’s suggestion—but 10 bills still left it out. Hawaii, Illinois, Oklahoma, and the ever-influential California all opted to retain sole authority over AI developers within their borders.
Asking states to do the job of Congress has failed. This is not to say that including a reciprocity provision in state legislation is a bad policy idea, per se, but that it is insufficient for the job at hand, and unlikely to be repeatable in other states across the country.
If the stars aligned and AI developers got 50 state laws with strong, matching harmonization clauses to a dependable, light-touch framework, then state-led AI regulation could work. Any states that decide to do it themselves should include such a clause if only to signal to developers that they are willing to cooperate if a national framework offers interstate consistency. But with America’s AI-powered future at stake, OpenAI’s call for regulatory harmony would be better sent to Congress.
Nate Karren is a policy analyst with the American Consumer Institute, a nonprofit education and research organization. For more information about the Institute, follow us on X @ConsumerPal.