America’s AI regulation debate is moving beyond whether powerful models should face rules and toward who gets to write them — as AI companies, states, Congress and the White House debate whether the country should operate under one national framework or multiple state standards.
WHAT’S HAPPENING
OpenAI is calling for mandatory national AI safety requirements based on model capabilities while continuing to support selected state-level legislation until Congress acts.
OpenAI Chief Global Affairs Officer Chris Lehane describes this state-first process as “reverse federalism” — states establishing similar safeguards that could eventually contribute to a national standard. OpenAI recently announced support for California measures covering independent safety assessments, AI-auditor standards, protections for young people and biological-risk safeguards.
At the same time, the White House is pursuing a national policy framework and has recommended that Congress preempt certain state AI laws while preserving traditional state powers in areas including consumer protection, fraud, children, zoning and government use of AI.
WHY IT MATTERS
The AI-policy debate is becoming less about a simple choice between regulation and no regulation.
The increasingly important question is:
Which level of government should establish the rules, and which rules should apply nationwide?
Supporters of a national standard argue that dozens of different state requirements could increase compliance costs, create conflicting obligations and make it harder for companies to operate AI systems nationally.
Supporters of continued state authority argue that states can respond more quickly to emerging risks and maintain stronger protections when Congress has not established comparable federal requirements.
Those competing approaches could determine how AI safety, accountability and innovation are governed across the United States.
WHO BENEFITS
A uniform federal framework could benefit companies operating nationwide by reducing the number of different regulatory systems they must navigate.
Smaller developers could also benefit if national standards prevent compliance requirements from becoming substantially different from state to state.
States, consumers and safety advocates could benefit from continued state authority where local laws address risks or protections not yet covered by federal legislation.
The ultimate impact depends heavily on what a federal framework actually requires and which state powers it preserves.
WHO LOSES
A broad federal preemption law could reduce the ability of states to impose AI-specific requirements beyond the national standard.
Conversely, maintaining significantly different state systems could increase compliance burdens for companies operating across the country.
Neither outcome automatically produces stronger or weaker AI oversight; the effect would depend on the substance, enforcement and scope of the rules adopted.
WHAT HAPPENS NEXT
Congress is considering several approaches.
Recent bipartisan Senate negotiations have included possible requirements for advanced AI developers to mitigate major known risks, conduct testing and potentially face federal intervention involving the release of particularly dangerous systems, with judicial review available. The negotiations have also considered whether some state laws should be preempted.
But Congress has not yet enacted a comprehensive national AI law.
That leaves states continuing to legislate, the executive branch pursuing its national policy, AI companies advocating for different forms of regulation and lawmakers negotiating the boundaries between federal and state authority.
The next phase of American AI policy may therefore be defined by a question larger than whether AI is regulated:
Who gets to write the rules — and which rules ultimately govern everyone?
