OpenAI has outlined a new policy stance it calls "reverse federalism" for U.S. AI regulation, arguing that state-level laws should serve as testing grounds that inform a future national framework, rather than waiting for a top-down federal mandate.
The 'Reverse Federalism' Playbook
In a policy announcement, OpenAI argues against a one-size-fits-all federal approach to AI regulation. Instead, it proposes a bottom-up model in which states experiment with their own rules, generating data on what governance approaches work before those lessons are consolidated into federal law. The company frames this as inverting the usual pattern of federal law setting the baseline that states later adapt.
States as AI Policy Laboratories
OpenAI points to California, Colorado, and Connecticut as states already legislating on AI transparency, bias audits, and risk assessments, citing this activity as evidence the state-led model is already underway.
According to OpenAI, the approach offers several advantages:
- Faster adaptation: states can move more quickly than Congress on emerging technology issues.
- Diverse solutions: different states can test different regulatory models, from risk-based frameworks to disclosure requirements.
- Reduced polarization: state-level debates are, in OpenAI's view, less partisan than national ones.
- Informed federal action: a track record of state laws could give future federal legislation a tested foundation.
Forging a National Consensus from Local Laws
OpenAI describes the end goal as a unified national standard distilled from the strongest state-level approaches, rather than a permanent patchwork. The company argues that as states refine their laws, common principles will emerge that can inform eventual federal legislation.
The proposal is a notable entry in the U.S. AI policy debate: rather than pushing for federal preemption, OpenAI is framing state experimentation as a feature rather than a problem to be solved. Whether that view holds depends on how easily the resulting patchwork of state rules can later be reconciled into coherent federal policy.