The United States is rapidly developing a state-by-state system for regulating artificial intelligence, and some of the governors creating those rules now say the country cannot continue that way indefinitely.
Maryland Gov. Wes Moore and Illinois Gov. JB Pritzker called on Congress and the federal government to establish national AI safeguards, arguing that a technology operating across state and national borders cannot realistically be governed through dozens of separate state frameworks.
Their comments came during national television interviews on September 20 and were followed almost immediately by new AI initiatives in both states.
Pritzker told ABC News that AI oversight “needs to be a national effort as well as an international endeavor” and added that “AI is not going to go away.”
Moore, speaking on CNN, made the same argument from the perspective of the states.
“Having 50 different guidelines is not the most productive thing for us to be doing and thinking about,” he said.
The comments capture one of the central problems emerging in U.S. AI policy.
Congress has not enacted a comprehensive national law governing the development and deployment of AI. States have responded by writing their own rules. Those state laws are now becoming substantial enough that developers and businesses may face materially different requirements depending on where they operate.
At the same time, the federal government has proposed its own national AI framework and has pushed back against state rules it considers too burdensome.
The debate is therefore becoming less about whether the United States needs a federal AI policy and more about what that policy should contain, how prescriptive it should be and which state protections should survive once national rules exist.
States Moved Because Congress Did Not
The current patchwork did not appear overnight.
State legislatures have spent several years moving into areas where federal AI legislation has remained incomplete.
In 2025 alone, all 50 states, Puerto Rico, the U.S. Virgin Islands and Washington, D.C., considered AI legislation. Thirty-eight states enacted or adopted roughly 100 measures, according to the National Conference of State Legislatures.
Those laws do not all regulate the same thing.
Some address deepfakes and election communications.
Others regulate automated employment decisions, healthcare applications, government use of AI, discrimination, children’s safety, synthetic intimate images or advanced AI models.
That experimentation has allowed states to respond quickly to specific problems.
It has also started to produce the fragmentation Moore is talking about.
A company deploying one AI system nationally may need to determine whether that system triggers disclosure requirements in one state, risk-management obligations in another and employment or consumer-protection rules somewhere else.
NCSL continues to maintain a national database of introduced and enacted AI legislation because the volume has become too large to treat state AI regulation as a collection of isolated bills.
Illinois Has Already Built Frontier AI Rules
Pritzker’s argument is particularly notable because Illinois is not waiting for Congress.
Illinois has enacted an AI Safety Measures Act that applies safeguards to advanced AI developers. According to Pritzker’s September 22 executive order, the law includes risk-mitigation frameworks, annual independent third-party audits, 72-hour reporting of critical safety incidents and public explanations of risks and mitigation measures.
Two days after calling for federal action on ABC, Pritzker created an Illinois Artificial Intelligence Cabinet.
The new body will bring together state officials and outside experts to examine AI safety, privacy, cybersecurity, critical infrastructure and governance.
Its mandate includes developing policies for responding to AI incidents, examining threats to infrastructure and assessing whether existing legal remedies are sufficient when AI causes harm. The Cabinet can also examine whether data-center incentives or state procurement should be conditioned on AI safety and accountability requirements.
Pritzker’s position is therefore not that Illinois should stop regulating AI.
It is that Illinois cannot solve a national and potentially international problem by itself.
Maryland Is Doing Much the Same Thing
Maryland is moving simultaneously.
On September 22, Moore announced a broader state AI framework focused on protecting residents, workers and children.
Among other measures, Maryland plans to develop legislative recommendations for frontier AI companies that could include safety frameworks, testing, independent third-party evaluations, whistleblower protections, public reporting and 72-hour incident reporting.
The state is also examining AI in employment, housing, lending, schools, gambling and children’s online services.
Maryland agencies are expected to clarify how existing civil-rights and consumer-protection rules apply when AI systems make or influence decisions. The state also plans to examine discriminatory tenant-screening algorithms, AI use in schools and protections involving chatbots used by children.
Moore’s administration expressly describes these efforts as necessary because of what it sees as insufficient federal guardrails.
Yet Moore also says state action is not the ideal end state.
That creates an unusual policy dynamic: governors are simultaneously building the regulatory patchwork and asking Congress to replace at least part of it with something more uniform.
The Federal Government Actually Wants Uniformity Too
There is an important complication to the narrative that Washington has simply ignored AI regulation.
The Trump administration has also argued that the United States needs a national AI framework.
Its preferred framework, however, looks different from what governors such as Moore and Pritzker are advocating.
In March 2026, the White House released legislative recommendations for a federal AI policy framework covering areas including children’s protections, energy, intellectual property, workforce issues and national AI policy.
The administration explicitly called for Congress to prevent a fragmented state regulatory system.
Its recommendations say federal legislation should preempt state AI laws that impose what the administration considers undue burdens, while preserving certain traditional areas of state authority, including generally applicable consumer-protection and child-safety laws, state government use of AI and zoning authority.
So both sides are now talking about national uniformity.
They differ substantially over what that uniformity should accomplish.
Federal Preemption May Become the Hardest Question
Any serious federal AI law will eventually have to address preemption.
In plain English: if Congress creates a federal standard, can states continue imposing their own AI requirements?
There are several possibilities.
Congress could establish a federal floor and allow states to impose stronger protections.
It could establish both a floor and a ceiling, preventing states from going further.
It could preempt state regulation only in specific areas, such as frontier-model safety, while preserving state authority over employment, housing, consumer protection and children’s privacy.
Or Congress could largely leave the existing state structure intact.
A bipartisan House AI task force previously noted that federal preemption has advantages and disadvantages. National rules can create consistency and reduce compliance burdens, while state regulation can allow experimentation and responses tailored to particular populations or industries.
This is not a technical detail.
It may determine whether federal legislation actually eliminates the “50 different guidelines” problem Moore identified.
The White House Has Already Challenged State AI Regulation
The federal-state conflict began before the current comments from Moore and Pritzker.
In December 2025, President Donald Trump issued an executive order directing the Attorney General to create an AI Litigation Task Force to challenge state AI laws that the administration believes conflict with national policy, federal law or constitutional requirements.
The order called for a minimally burdensome national framework intended to support U.S. AI development and competitiveness.
The administration’s March 2026 legislative recommendations later stated that states should not be permitted to regulate certain aspects of AI development that the White House views as inherently interstate and connected to national security and foreign policy.
That is materially different from the approach Illinois is taking with developer audits, risk frameworks and incident reporting.
It also differs from Maryland’s planned frontier-model legislation.
The result is a federalism dispute sitting underneath the broader AI safety debate.
The Question Is Not Simply Regulation Versus No Regulation
That distinction gets lost easily.
The current fight is not cleanly divided between officials who want AI regulation and officials who want none.
The Trump administration has enacted measures involving advanced AI and national security and has proposed federal legislation. In June 2026, for example, a White House executive order established federal measures focused on advanced AI innovation and security while maintaining the administration’s preference for avoiding what it describes as overly burdensome regulation.
Moore and Pritzker want stronger safeguards and a larger regulatory role for government.
The administration has emphasized innovation, U.S. competitiveness and limiting state-level restrictions.
Those are competing models of national AI governance, not simply the presence or absence of a federal policy.
Why Governors See AI Differently From Most State Issues
AI creates several problems for ordinary state regulation.
The developer may be located in California.
The data center may be in Texas.
The model might serve a company in Illinois.
The affected consumer might live in Maryland.
The underlying model weights may be deployed globally.
An AI incident involving cybersecurity or biological capabilities could cross national borders almost immediately.
That is one reason Pritzker argues that frontier AI safety cannot stop at state borders.
Illinois can require certain companies within its jurisdiction to perform audits or report incidents.
It cannot establish international standards for frontier-model testing or coordinate U.S. policy with foreign governments.
Those functions require federal involvement.
But States Regulate Areas Washington Traditionally Does Not Control Alone
The opposite problem is equally real.
Many AI harms occur in areas traditionally regulated heavily by states.
Employment is one example.
Housing is another.
Insurance, education, professional licensing and consumer protection all have significant state-law components.
An AI system that determines whether someone receives an apartment may present a different regulatory problem from a frontier model capable of sophisticated cyber operations.
That makes one universal AI statute difficult to design.
A federal law focused primarily on model developers may do little about a discriminatory tenant-screening product.
A state employment algorithm law may do little about an advanced model’s cybersecurity capabilities.
A future national framework may therefore need layers rather than one rule covering everything called artificial intelligence.
Businesses Are Already Living With the Patchwork
For companies adopting AI, this is not an abstract constitutional debate.
A national employer using automated hiring tools already has to consider state and local employment rules.
A company building consumer-facing AI needs to monitor state privacy, discrimination and children’s safety requirements.
Advanced model developers increasingly face state-specific safety and transparency obligations.
Organizations can therefore find themselves maintaining an AI compliance matrix much like the privacy-law matrices companies built after the GDPR and the first wave of comprehensive U.S. state privacy laws.
If that sounds familiar to privacy professionals, it should.
The United States followed a similar path with consumer privacy.
California moved first with broad privacy legislation. Other states adopted different models. Congress repeatedly considered federal privacy legislation without enacting a comprehensive nationwide replacement.
Businesses now comply with a growing collection of state privacy regimes.
AI could repeat that pattern at a much faster pace.
What Would a Federal AI Law Actually Need to Resolve?
A meaningful national framework would need to answer questions that state legislation is currently answering inconsistently:
- whether frontier AI developers must perform independent safety evaluations;
- which incidents must be reported and how quickly;
- whether companies must publish risk-management frameworks;
- what protections apply when AI affects employment, housing, credit or healthcare;
- how children should be protected when interacting with AI systems;
- what transparency consumers should receive when dealing with automated systems;
- how whistleblowers inside AI companies are protected;
- which agency or agencies enforce the rules; and
- how much authority states retain after a federal law takes effect.
The last question could prove harder than the others.
State Experimentation Is Producing the Rules Congress Could Eventually Borrow
There is another reason the current state activity matters even if Congress eventually acts.
States are effectively testing different models of AI regulation.
One state might require third-party audits.
Another might focus on incident reporting.
Another may prioritize discriminatory automated decisions.
Others may concentrate on children, deepfakes or government procurement.
Congress can observe which mechanisms prove workable before incorporating some of them into a national standard.
This is one of the traditional arguments for allowing states to operate as policy laboratories.
The downside is that companies must comply with the experiments while they are happening.
The Governors’ Calls Arrive as Their Own AI Programs Expand
The timing of the Moore and Pritzker comments makes the current moment particularly interesting.
Moore spoke on September 20 about the inefficiency of 50 different AI guidelines.
Two days later, Maryland announced a broad state AI agenda that could produce additional frontier-AI, employment, housing, children’s safety and education requirements.
Pritzker used the same weekend to argue that AI cannot be controlled state by state.
On September 22, he established a state AI Cabinet tasked with examining still more Illinois AI safeguards.
That is not necessarily contradictory.
It illustrates the problem they are describing.
Governors cannot simply stop governing while they wait for Congress.
But every new state initiative adds another piece to the regulatory system they say eventually needs national coordination.
Federal Rules Would Not Necessarily Mean Weaker or Stronger Rules
For privacy and compliance professionals, it is worth resisting the assumption that “federal regulation” automatically means one policy outcome.
A national law could create stronger mandatory safety requirements than some states currently impose.
It could also preempt state requirements and replace them with a less prescriptive federal standard.
It could preserve state consumer protections while centralizing frontier-model oversight.
Or Congress could create a federal baseline and allow states to go further.
Those choices matter much more than whether the final statute has “federal AI framework” in its title.
The Next AI Fight May Be Over Who Gets to Write the Rules
Moore and Pritzker’s comments show how quickly the AI policy debate is evolving.
States initially stepped in because they believed Washington was moving too slowly.
Now some of the same state leaders say the resulting patchwork cannot be the permanent solution.
The federal government agrees that uniformity has value but has proposed a model that would restrict some state regulation in the name of innovation and national competitiveness.
That puts the United States on a path toward a larger argument over regulatory authority.
How much AI safety belongs at the federal level?
Which consumer and civil-rights protections should states continue controlling?
Should one national rule govern frontier developers?
And if federal legislation arrives, does it establish a minimum standard or prevent states from imposing anything stronger?
Those questions are increasingly important because the state regulatory system is no longer hypothetical.
Illinois already has frontier AI safeguards.
Maryland is preparing more.
California, Texas and other states have adopted their own approaches.
And state legislatures around the country continue introducing AI bills.
Moore’s concern about “50 different guidelines” is therefore not a warning about something that might happen years from now.
The fifty-state AI experiment is already underway.
The question is whether Congress eventually turns those experiments into a national framework or allows the American AI regulatory map to continue developing state by state.