On July 23, the Trump administration released its long-anticipated AI Action Plan aimed at accelerating innovation and “winning the AI race” for global supremacy. A centerpiece of the plan is deregulation: removing what the White House views as domestic barriers to AI development. While the stated policy applies nationwide, its most significant impact may fall on states, where the vast majority of AI regulations currently exist.
The plan’s approach to state regulation inverts the traditional model of cooperative federalism. Most commonly, Congress uses its spending power in cooperative programs to entice states to partner with the national government and adopt federal standards. The AI Action Plan, by contrast, would induce states to idle in the face of mounting AI-related social concerns. Instead of building a cooperative state-federal regulatory regime, the White House hopes to create and maintain a regulatory vacuum at the state level by executive fiat.
Two provisions are particularly consequential. One directs the Office of Management and Budget (OMB) to restrict “AI-related” funding to states that adopt what the administration considers “burdensome” or “unduly restrictive” AI regulations. The second directs the Federal Communications Commission (FCC) to examine whether state AI regulations “interfere” with federal authority, potentially laying the groundwork for administrative preemption.
This article argues that the plan’s approach to state regulation is not just a departure from traditional cooperative federalism—it also raises serious constitutional concerns. First, the AI Action Plan undermines the separation of powers by imposing funding conditions that Congress has never authorized. Second, it stretches administrative law limits by inviting federal agencies, particularly the FCC, to preempt state AI regulations despite lacking a clear statutory basis to do so. Finally, the plan threatens core principles of federalism by attaching funding restrictions that are ambiguous in scope, potentially unrelated to the federal programs they affect, and coercive in their impact on states’ ability to legislate for the protection of their residents. Each of these legal fault lines is discussed in turn.
Separation of Powers
The Constitution expressly vests the spending power with Congress, not the executive branch. Most pertinent here, Congress has appropriated funds for a wide array of programs, from infrastructure development to healthcare, transportation, education, law enforcement, and more. Each program has its own statutory purposes and conditions. The AI Action Plan proposes a new, executive-created condition: namely, that states forgo AI oversight that the executive branch deems “burdensome” or “unduly restrictive to innovation.”
What qualifies to trigger these restrictions is unclear, but that ambiguity is by design. It will have the effect of chilling some state action ex ante, and it allows the administration to punish states ex post whenever and wherever the administration disfavors a state’s regulatory approach.
However, the Constitution and federal law prohibit the executive branch from unilaterally imposing new funding conditions or withholding funds that Congress appropriates. Indeed, only weeks prior, the Senate considered and overwhelmingly rejected a similar spending condition on state AI regulation that was appended to the One Big Beautiful Act. That legislative defeat leaves no doubt about Congress’s intentions, and no room for the president to impose similar spending conditions unilaterally.
Administrative Law
Beyond the direct separation-of-powers conflict, the AI Action Plan’s strategy also relies on a vision of administrative power that is profoundly out of step with the current legal landscape. As previously noted, the plan directs the FCC to “identify state AI regulations that interfere with the FCC’s mandate under the Communications Act.” The clear subtext of this directive is to lay the foundation for federal preemption through the administrative process.
While the Supreme Court has long held that agencies may preempt state law, there are important limits. Most pertinent here, only valid federal law can preempt state law. The FCC will be hard-pressed to meet this burden under its existing statutory authority. Nothing in the Communications Act confers FCC authority to regulate AI. Nor has Congress delegated authority to the FCC to preempt state AI regulations. Any such claim would, at best, rely on statutory ambiguities or vague generalities—neither of which will suffice.
In the past, the FCC might have received judicial deference for broad interpretations of statutory authority. But those days are over. In Loper Bright v. Raimondo (2024), the Supreme Court overruled the seminal Chevron doctrine, which required courts to defer to an agency’s “reasonable” interpretation of ambiguous statutes. Post-Chevron, no such deference is due.
What’s more, any effort by the FCC to preempt state law will be scrutinized under the major questions doctrine, which prevents agencies from regulating issues of “vast economic and political significance” based on vague or general statutory language. Instead, the agency must be able to point to “clear congressional authorization” for the specific action at issue.
The preemption of state AI regulation could easily qualify as a “major question,” although it would depend on the substance and scope of the federal policy at issue. AI is a general-purpose technology that impacts nearly every sector of the economy and society, and states are actively legislating in areas ranging from consumer protection and employment discrimination to public safety and civil rights. There simply is no clear statutory authorization for the FCC to become the nation’s primary AI regulator, much less the arbiter of how states exercise their sovereign authority to protect the health, safety, and civil liberties of their residents.
The Court’s overruling of Chevron and its embrace of the major questions doctrine are related phenomena. Both link to a vision of separation of powers that positions courts as counterweights to executive overreach. Although seldom appreciated, these administrative law developments implicate federalism too. Together, they create substantial barriers to unauthorized executive action, leaving policy gaps that states can fill.
Federalism
Finally, the AI Action Plan’s proposal conflicts with core federalism principles. To pass constitutional muster, federally imposed spending conditions must be (i) unambiguous, (ii) germane, and (iii) non-coercive. The plan’s proposal to “cut off AI-related funding” is vulnerable on all three fronts.
To start, the terms of the plan are irredeemably ambiguous. The plan deliberately avoids defining or describing “burdensome” state regulations that would trigger federal withholding. Moreover, “AI-related” funding could range from R&D programs to education technology funds, to infrastructure monies for broadband, to law enforcement and beyond. The resulting ambiguities may leave states unable to make informed decisions, which is precisely why the Court requires spending conditions to be clear and unambiguous—as a check on federal power vis-à-vis the states.
The breadth of the funding condition may also run afoul of the germaneness test, which requires that spending conditions must be reasonably related to the federal interest in the particular national projects being funded. That nexus cannot be gleaned from the text of the AI Action Plan, which is exceptionally broad and untethered. It threatens to withhold a wide range of discretionary federal funding based on a state’s AI regulatory regime, without specifying which funding streams might be implicated or in which substantive domains. Thus, for example, a state that regulates the use of AI in employment contexts might lose federal grants from the Department of Transportation to optimize traffic flow. Of course, the federal government might apply the condition more narrowly. The point is that the AI Action Plan itself contains no such limits.
Lastly, federal spending conditions cannot be coercive. Again, depending on implementation, the plan’s funding cutoff could be deemed coercive based on the amount of funds involved and whether pre-existing funding programs would be cut. Without further guidance, the scope of AI-related funding is amorphous. If aggressively implemented to encompass federal grants for science, technology, education, infrastructure, and so forth, it could cross the line from permissible incentive to unconstitutional compulsion. Nor does the plan rule out clawing back funds retrospectively or attaching new conditions on existing funding streams. Suffice to say, such retrospective applications would undermine a state’s reliance interests and heighten constitutional concerns.
* * *
Sidelining the states may seem like a politically expedient approach to AI governance. But it is fundamentally incompatible with our constitutional democracy. The AI Action Plan repeatedly invokes the importance of instilling “American values” into AI systems and seeks to cut funding to states that regulate AI. Yet there is a glaring contradiction in these ambitions. The plan’s approach to AI federalism could suppress the very democratic and legislative processes through which Americans, at the state level, express their diverse values and preferences.
Ultimately, the Trump administration’s approach to AI as articulated in the Action Plan is starkly at odds with legal precepts across three pillars of public law: the separation of powers between the executive and legislative branches, the doctrines of administrative law, and the foundational principles of federalism that delineate the authority of the national government relative to the states. The constitutional analysis also reveals the broader stakes involved. The AI funding conditions test the President’s power to achieve what Congress overwhelmingly rejected. But unlike other (ongoing) tests that pit executive power against Congress or the courts, the federalism components of the AI Action Plan pit executive power against sovereign states. If the approach survives constitutional scrutiny, it will erode vital and enduring checks on presidential authority.
