America is at a fork in the road. Artificial intelligence is poised to define the twenty-first century. Whoever wins the AI race has the chance to achieve unprecedented prosperity, lead the globe in growth, and live in secure peace. The potential is limitless.
For my AI-skeptical friends, there is no ducking this fight. Artificial intelligence is coming whether we like it or not. China is not going to wait—they are already playing to win. We have no choice but to do the same.
To outperform our competitors, the United States needs uniform legal rules that facilitate the development of revolutionary technology. We don’t have that right now. Rather than a national framework that governs AI development, American builders face an increasingly chaotic patchwork of state-by-state regulations that encumber already difficult engineering work with endless compliance requirements. This is a recipe for stagnation—look at Europe, not for world-historical innovation.
Blessedly, President Trump gets it. At the “Winning the AI Race” summit on July 23, he told the crowd that winning the global AI race will depend on “common sense” principles. AI cannot thrive, President Trump explained, with “lowest common denominator” regulation where states out-compete each other in stifling growth.
The President is correct. American companies can’t create 50 products for 50 states. Unless the federal government provides a uniform standard, the most anti-AI state legislatures will dictate the rules for everyone.
Case in point: SB-53, which California Governor Newsom signed into law last month, establishes what the state calls “guardrails” for “responsible and ethical AI.” What that means in practice is that the California Attorney General will determine what constitutes a “health and safety risk” when American innovators build AI tools. Senator Ted Cruz called SB-53 what it is—“government micromanagement”—rightly noting that it “puts California bureaucrats in charge of policing artificial intelligence.” Most people have decided against living in California. Those who have made that choice shouldn’t nonetheless submit to California’s legislature making national AI policy.
Building AI technology—such as training large language models and developing operational parameters—is a fundamental component of economic growth and global competitiveness. It needs to be regulated at the national level. Congress should therefore enact legislation preempting the expanding patchwork of state rules that govern the construction and design of AI tools while the federal government develops a uniform national standard.
This does not mean preempting everything that involves artificial intelligence. Congress should explicitly encourage states to pass legislation that facilitates AI deployment, reduces bureaucratic hurdles, and eliminates conflicting design mandates. There is also a role for states in regulating AI outputs, the harms that AI tools cause to consumers and users, and the actions that AI tools take after they’ve been built. Nor should Congress prevent states from passing, for example, age-verification and parental consent laws that protect minors from harm online. State laws governing consumer protection—a quintessential state issue—should also not be affected by a federal pause on AI legislation. But we cannot afford to allow states to dictate how AI systems are built.
Consider the rules for building cars. Every state has its own rules about how cars are driven, when and how automakers are liable for product defects that hurt consumers, or what happens when design defects cause dangerous accidents.
But that doesn’t mean we should let California set its own fuel emissions standards. The last thing we want is California leveraging its population to force the production of “green” cars that nobody wants and that the nation did not vote to demand. Indeed, Congress and President Trump recently used the Congressional Review Act to evaporate a Biden-era waiver of federal preemption that would have permitted California to do just that. That was right for cars. It is right for AI too.
Builders cannot build if they need to consult lawyers about 50 different rules for every aspect of their construction. Builders need one set of rules to govern the basics—how tools are trained, what data their engineers use to train the model, and so forth. Without federal preemption, one state could prohibit training a large language model to embrace woke tropes about gender ideology or racial equity, while another state requires that very same training. It is a recipe for impossible mandates and ultimately European-style stagnation.
Industry experts agree that without a national AI policy, we’re going to end up with “50 shades of red tape.” The Abundance Institute, the International Center for Law and Economics, and R Street concur. Remarkably, more than 1,000 AI-related bills have been introduced in 2025. On those facts, the case for federal preemption makes itself.
As President Trump declared, we must let AI “thrive” without an avalanche of “foolish rules.” That doesn’t mean no rules. It means smart rules—i.e., “common sense” uniform federal rules that enable the most intelligent, most innovative people in the world—Americans—to stay first in artificial intelligence for the next hundred years and beyond.
The United States wins when we let our people build. That means getting governments out of the way. The internet flourished under federal leadership, not state-by-state micromanagement. AI demands the same approach. If we want to win the defining race of the century—and we have more than enough talent to do it—Congress needs to act immediately on federal preemption.
China isn’t waiting. Why are we?
James Burnham is a former senior White House and DOJ official, the former General Counsel at the Department of Government Efficiency (DOGE), and a former clerk to Supreme Court Justice Neil Gorsuch. He founded the AI Innovation Council to bring together leaders across law, technology, business, and defense with one mission: to promote America First AI.
