The National Governors Association AI principles, released Sunday, call on Congress to write a federal AI framework that leaves room for state laws. The bipartisan group wants the incoming Congress to take up the issue as lawmakers start the next session, which puts the federal-versus-state fight over AI rules at the center of the agenda.

The NGA's press release lays out a plan that divides the work. Washington would address national and international risks as AI is developed. States would keep the flexibility to protect people and regulate how AI is used once it reaches the public.

What the governors are asking Congress to do

The principles come from the bipartisan Governors Council to Secure American AI Dominance, which NGA Chair Maryland Gov. Wes Moore leads. NGA Vice Chair Indiana Gov. Mike Braun is involved, as are Governors Josh Stein of North Carolina and Spencer Cox of Utah. The association counts 55 members, covering the governors of the 50 states plus five territories and commonwealths, according to the NGA's entry on Wikipedia.

In the release, Moore made the request directly: "we are calling on Congress to work with us to develop a federal framework that strengthens American competitiveness while protecting the public from harm." The governors say they want a shared national approach, more consistency across state frameworks, and useful input for a Congress that is writing its own AI framework that respects federalism.

Moore has framed the effort as a response to Washington's inaction. Reporting on the group's formation said states "cannot just sit on our hands" while federal leadership is missing. Moore has also argued that state efforts "cannot happen in the absence of federal leadership" in a separate Yahoo News interview.

How the proposed split divides AI oversight

The core idea is a division of labor by stage. Washington takes the risks that arise as models are built and that cross state and national borders. States take the harms that arise when a company, school or government agency puts an AI system to work.

Level Role under the NGA proposal
Washington National and international risks as AI is developed
States Protecting people and regulating how AI is used once deployed

Beyond the split, Startup Fortune's summary of the recommendations says the NGA wants Congress to cut duplicative or conflicting rules, tie safeguards to evidence of actual risk, and leave room for smaller firms to compete. Those three demands point in the same direction. They ask for fewer overlapping rules, more proof before new restrictions, and a path for startups that cannot absorb the cost of 50 different compliance regimes.

In practice, the line would sort AI questions by where they arise. A question about how a bank, hospital or employer uses an AI tool would stay closer to home, with state regulators. A question about risks that spill across state lines or borders would fall to federal officials. That distinction is the proposal's most important feature, and it is also where the friction begins.

The NGA materials reviewed for this story do not spell out specific recommendations on safety testing, child protection or enforcement. Those details will matter once the principles turn into bill text, and readers should check the full document for them.

Why the White House preemption question is the sticking point

The fight over preemption predates these principles. In March, a White House legislative framework asked Congress to preempt state AI laws that impose "undue burdens." The same framework would leave child-protection, consumer-protection, fraud, zoning and government-procurement rules alone, according to Startup Fortune's account.

Startup Fortune reports that the NGA plan clashes with the White House on preemption. It frames the plan as an effort to keep state laws in place under a single federal rulebook. The NGA excerpts reviewed for this story do not name the White House, so the clash is Startup Fortune's characterization rather than language from the association. The sources also do not say whether the White House has responded to the principles.

The phrase "undue burdens" is the detail to watch. The sources do not define the threshold, which leaves the reach of any federal preemption unsettled. A narrow reading could displace a small set of state rules, while a broad one could sweep in many of the AI laws states have already passed.

Who gains and who loses if Congress acts

Consumers have the most at stake in the details. Under the March framework's carve-outs, state child-protection and consumer-protection rules would stay in place. Rules outside those categories could be displaced if Congress acts broadly. Residents of states that have already written their own AI rules would feel that change first.

Businesses, especially smaller ones, face a different calculation. A single federal standard could simplify compliance for companies that operate across state lines. The NGA says it wants to avoid duplication and keep room for smaller competitors, which suggests the governors see a federal floor as a way to reduce costs without wiping out local protections.

Governors' leverage is uneven. According to SAN's reporting, governors have a better chance of exerting regulation in states where AI companies are headquartered or where data centers are being built. Some commentators question how much governors can accomplish on AI regulation at all.

The governors are not all on the same page, either. Republicans are acting at the state level too. Utah Gov. Spencer Cox and Texas Gov. Greg Abbott have both cracked down on data center development, according to Yahoo News. That shows the bipartisan coalition is working on a shared goal of federal coordination, even as its members take different paths on specific issues.

State action is also moving on its own track. California Gov. Gavin Newsom has proposed an AI kill switch and 48-hour incident reports, an example of the state-level activity now underway that the federal debate could affect.

What readers can do now

Nothing in these principles changes the law today. They are recommendations to Congress, and no bill carrying them had been confirmed in the sources reviewed. Readers who want a voice in the outcome have a few concrete options.

  • Read the NGA's release and the full principles document for specifics on testing, child protection and enforcement, which the summaries do not detail.
  • Look up your state legislators' pending AI bills and note which ones cover child safety or consumer protection, the categories the White House framework would leave to states.
  • Ask your U.S. representative and senators where they stand on preempting state AI laws, and whether they would keep state rules on child protection and consumer protection intact.
  • Track bill numbers once legislation is introduced in the next session. Bill text is the first point at which any of these ideas could change what you are allowed to do or what companies must do.

What to watch in the next session

The next signal is whether any member of Congress introduces legislation that carries these principles, and what that text says about "undue burdens." A bill that defines the threshold would tell states far more about their room to act than any press release. The White House's response will matter as well, and the sources do not show one yet.

Federal pressure is building from other directions too. Senators have urged the White House to act on AI agent cyber threats, a reminder that Washington has its own reasons to set national rules. Whether that concern pushes Congress toward broader preemption or toward a narrower federal floor is an open question.

The governors have put a specific division of labor on the table. Whether Congress accepts it, rewrites it or sets it aside will determine how much of today's state AI law survives the next session.