Commentary. The political argument over artificial-intelligence regulation has moved past the question of whether rules are necessary. The more consequential question is who gets to write them—and whether federal legislation will establish a minimum standard or quietly prevent anyone else from doing more.
Our view is straightforward: Congress should create a national floor for AI safety and accountability, not a ceiling that disables states. A single federal framework could reduce uncertainty for companies and give the public enforceable rights. But a sweeping pre-emption clause would repeat a familiar Washington mistake: treating regulatory uniformity as more important than learning what works.
The issue is becoming urgent. A proposed congressional bill is reportedly being revised to narrow language that would pre-empt state AI laws, after Democrats objected to blocking states from acting and technology companies warned that complying with divergent rules would be unworkable.[1] The debate follows a Senate vote in July that rejected a broad provision preventing states from regulating AI.[3]
Why a national standard matters
Companies building and deploying powerful models operate across state lines. A patchwork of definitions, testing requirements and disclosure rules could impose significant costs, especially on smaller firms that cannot maintain separate compliance systems. Industry advocates have a legitimate point: inconsistent obligations can slow beneficial applications in medicine, education and scientific research while giving the largest companies an advantage because they can afford armies of lawyers.
National rules could also clarify responsibility. Proposals discussed publicly include risk assessments, transparency requirements, testing and clear accountability for developers.[2] Those measures would be more useful than broad promises of voluntary restraint, particularly when systems can generate convincing falsehoods, expose personal information or make high-stakes decisions at scale.
A federal baseline would also make enforcement less arbitrary. Consumers should not receive radically different protections simply because an automated hiring tool, financial service or health application is offered across a state border.
Why pre-emption is the wrong bargain
Uniformity becomes dangerous when it is achieved by forbidding experimentation. State governments are often where new risks first become visible, and their rules can function as laboratories for national policy. If lawmakers in one state identify a harmful practice sooner than federal agencies do, residents should not have to wait for Congress to catch up.
The strongest case against a blanket prohibition is practical, not ideological. AI is changing faster than legislation. A statute that locks in today’s assumptions could leave regulators unable to address tomorrow’s abuses. The recent congressional debate reflects that tension: some Republicans and industry groups favor pre-emption, while Democrats have largely resisted blocking state action.[1]
There is a counterargument worth taking seriously. State-by-state regulation can produce contradictory obligations and invite forum shopping. Smaller developers could abandon markets rather than navigate legal uncertainty. And excessive compliance costs might entrench the biggest incumbents—the very companies policymakers say they want to discipline.
That concern supports coordination, however, not surrender. Congress can define core obligations, establish interoperable reporting standards and create a federal regulator or enforcement mechanism. It can expressly pre-empt only rules that directly conflict with those requirements, while preserving stronger protections for privacy, children, discrimination and consumer fraud.
The test should be accountability
Some voices now call for pausing advanced AI development or even banning artificial superintelligence.[9] Such proposals reflect genuine anxiety, but they risk collapsing distinct problems into one dramatic solution. The immediate policy challenge is not merely a hypothetical future machine; it is the deployment of existing systems whose failures already affect people and institutions.
Others argue that catastrophic-risk warnings distract from present harms and could produce rules that favor established technology companies.[6] That criticism should shape legislation. Safety policy must address discrimination, deception, labor disruption, security and energy use—not only distant scenarios. It must also prevent compliance regimes from becoming an expensive moat around the industry’s largest firms.
The right standard is therefore neither laissez-faire nor blanket prohibition. It is enforceable accountability proportionate to risk, transparent enough for independent scrutiny and flexible enough to evolve. Federal lawmakers should set that floor, fund competent oversight and leave states room to respond when evidence shows the floor is too low.
The AI race does not justify a race to the bottom. Innovation deserves predictable rules; the public deserves the ability to demand better ones. A national framework that preserves both would be harder to negotiate than sweeping pre-emption—but far more worthy of the technology it seeks to govern.