How a Federal Executive Order Rewrites the Rules of US AI Federalism

The United States is currently navigating a profound constitutional standoff over who holds the authority to govern artificial intelligence. On one side of the divide, the federal executive branch is pushing aggressively for a centralized, unified national framework, culminating in Executive Order 14365, which seeks to preempt state-level AI regulations.

On the other side, individual states are rapidly enacting their own safety, transparency, and consumer protection mandates, arguing that they cannot wait for a gridlocked Congress to address the immediate risks of frontier models. This clash between federal supremacy and state sovereignty has transformed AI policy from a purely technical debate into a fundamental test of American federalism.[1][2]

The mechanism of federal action has increasingly relied on administrative leverage rather than direct legislative preemption. Because Congress has yet to pass a comprehensive AI statute—despite years of high-profile hearings and sweeping announcements—the executive branch has deployed a strategy that legal scholars term 'managed federalism.'

The December 2025 executive order established an AI Litigation Task Force, directing the Department of Justice to challenge state laws that conflict with the administration's deregulatory posture. More controversially, the order linked federal infrastructure support—specifically threatening to withhold portions of the $42 billion Broadband Equity Access and Deployment program—to a state's willingness to align with the national AI framework, effectively using financial coercion to achieve what legislation has not.[1][5]

Despite these federal pressures, states have continued to act as the primary laboratories for AI governance, shipping actual legislation while Washington debates frameworks. California's landmark SB 53, which went into effect in January 2026, requires developers of frontier AI models to publicly disclose their safety and security protocols—a tangible mandate that goes further than even the European Union's AI Act.

Other states have followed suit, passing laws that address algorithmic discrimination in hiring, deepfakes in elections, and data privacy. Proponents of this decentralized approach argue that state legislatures are uniquely positioned to test regulatory models, generating vital empirical data on what policies effectively mitigate harm without stifling innovation.[2][3]

The constitutional limits of executive preemption remain a significant hurdle for the centralization effort. Legal experts consistently note that an executive order cannot unilaterally displace state law; under the Tenth Amendment and established Supreme Court precedent, only Congress possesses the authority to formally preempt state police powers.

Consequently, the federal government's reliance on conditional funding and strategic litigation represents an attempt to alter the political calculus for state lawmakers rather than a formal reallocation of legislative authority. This indirect approach has already shown some deterrent effect, causing proposed AI bills in states like Utah to stall amid opposition from federal officials and industry stakeholders.[1][2]

The push for a unified federal standard is heavily supported by the technology industry, which frequently frames state regulation as an existential threat to innovation. Industry advocates argue that navigating conflicting state laws disproportionately harms startups and smaller developers, who lack the legal resources to ensure nationwide compliance.

Furthermore, proponents of centralization contend that imposing a 'minimally burdensome' national standard is essential for maintaining American technological dominance. However, this rhetoric often conflates the genuine difficulty of compliance with a broader desire to avoid strict safety mandates, particularly as global competitors like China accelerate their own state-backed AI initiatives without the friction of internal jurisdictional disputes.[3][6]

Conversely, defenders of AI federalism argue that total centralization is not only politically unlikely but structurally unworkable. The actual deployment of artificial intelligence occurs within specific application layers—such as healthcare, transportation, and employment—where states already hold entrenched, non-preemptable regulatory authority.

For example, 22 states act as the primary regulators of occupational health and safety, giving them direct oversight over how AI is integrated into industrial automation and workplace management. Stripping states of their ability to govern these domains would require dismantling decades of established administrative law, a reality often ignored in sweeping federal policy announcements.[4]

The legislative battle over preemption reached a critical juncture with recent congressional proposals aimed at imposing a 10-year moratorium on state and local AI regulation. While this proposal ultimately failed to secure sufficient backing, it highlighted the intense desire among some federal lawmakers to clear the regulatory landscape for AI developers. Critics of the moratorium argued that prohibiting state action for a decade would create a dangerous governance vacuum, leaving citizens exposed to algorithmic harms while the federal government struggles to build consensus on a comprehensive national framework.[2]

As the standoff between the executive branch and state legislatures intensifies, the courts will increasingly serve as the arbiters of AI federalism. The newly formed AI Litigation Task Force is expected to initiate legal challenges against the most stringent state laws, testing the boundaries of the Commerce Clause and the limits of federal funding conditions.

Until a definitive judicial ruling or a comprehensive congressional statute emerges, the governance of artificial intelligence in the United States will remain a complex, contested negotiation between the desire for national uniformity and the democratic imperative of local control.[1][3][5]