When autonomous driving startup PerceptIn set out to navigate America’s artificial intelligence regulations, it budgeted $10,000 for compliance. The actual bill exceeded $344,000 per deployment project, more than double the company’s research and development costs. Two months ago, PerceptIn went out of business.
Last year, states introduced more than 1,200 AI-related bills, with at least 145 becoming law, creating contradictory requirements that multiply compliance burdens. Each jurisdiction defines “artificial intelligence,” “high-risk systems,” and “consequential decisions” differently, forcing companies to analyze identical technology under multiple incompatible frameworks. A hiring tool that satisfies California’s four-year recordkeeping and anti-bias testing requirements must also meet Colorado’s separate impact assessment mandates and New York City’s independent bias audit regime with distinct notice requirements. Each jurisdiction defines core terms differently, forcing companies to analyze identical systems under multiple incompatible frameworks.
Industry estimates suggest compliance costs add approximately 17 percent overhead to AI system expenses. For small businesses, California’s privacy and cybersecurity requirements alone impose nearly $16,000 in annual compliance costs. But these figures understate the true burden because they treat compliance as a variable cost that scales with company size. The reality is far worse.
Harvard Kennedy School researchers identified what they termed a “compliance trap” in which regulatory costs consume resources faster than startups can generate revenue. Their analysis found that a 200 percent increase in fixed compliance costs transforms a startup’s operating margin from positive 13 percent to negative 7 percent. That’s not a rounding error, it’s the difference between survival and bankruptcy. A three-person team building an employment screening tool faces the same baseline compliance obligations as a thousand-person enterprise in many jurisdictions, but without the revenue base or legal infrastructure to absorb those costs.
This dynamic hands an enormous competitive advantage to the very companies state regulations purport to constrain. Incumbent tech giants maintain compliance departments that dwarf entire startups. They can afford multi-jurisdictional legal teams, custom bias auditing frameworks, and the political relationships necessary to shape emerging requirements. For them, state AI fragmentation represents a manageable cost of doing business. For startups, however, it represents an insurmountable barrier to entry.
The strategic implications are staggering. While American entrepreneurs waste engineering talent on contradictory compliance regimes, Chinese AI companies operate under a unified national framework. Beijing’s approach is hardly a model of light-touch regulation, but it offers what America’s patchwork cannot: coherent rules that don’t change at state lines. When compliance costs exceed development budgets, innovation doesn’t slow—it stops altogether or moves to foreign jurisdictions where the rules are clear.
The White House recognized this competitive danger in a December 2025 executive order criticizing the “patchwork of 50 different regulatory regimes” and directing the U.S. Department of Justice to establish an AI Litigation Task Force to challenge state laws that obstruct national AI policy. The order represents a necessary first step, but executive action alone cannot solve a problem rooted in legislative fragmentation. Congress must act.
Federal preemption legislation would establish uniform national standards for AI systems while preserving states’ ability to enforce general consumer protection laws. This isn’t a radical concept. Federal frameworks already govern aviation safety, pharmaceutical approvals, and telecommunications—industries where state-by-state variation would create identical chaos. Safe harbor provisions could protect companies that implement reasonable bias testing and impact assessments from liability, creating incentives for responsible development without imposing contradictory mandates.
The current trajectory is unsustainable. Every month that passes with this regulatory chaos intact represents another month of American innovation surrendered to better-organized competitors. State legislators designed their AI frameworks to constrain Big Tech’s power. Instead, they’ve built a moat around incumbents while crushing the startups that might challenge them. The compliance trap isn’t protecting consumers. It’s protecting monopolies and handing a competitive advantage to foreign adversaries. That’s not regulation, it’s self-sabotage.
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.
This story was originally featured on Fortune.com
