Version 3

The Trump administration has signed an executive order preempting state AI laws and centralizing federal oversight [1][2][11], while simultaneously declining to impose substantive federal AI safety regulations — a combination that legal analysts and policy observers are describing as an emerging 'AI preemption war' [12]. California has countersigned: Governor Newsom's AI workforce executive order is now formally signed [7][8], and a separate California AI certification standards order (N-5-26) is also in effect [9]. The federal preemption mechanism is drawing scrutiny from law firms, constitutional scholars, and policy groups across the political spectrum [13][14][6][12].

The combination of active federal preemption and federal inaction on substantive AI standards creates what analysts are calling a governance vacuum: states are blocked from protecting workers and consumers, while no federal framework replaces those protections [13][6][12]. With California defending its workforce and certification orders and the legal community flagging constitutional limits on executive preemption, litigation appears likely — and the outcome will determine whether any level of US government can govern AI's economic and safety impacts in the near term.

American AI governance in mid-2026 is defined by a structural conflict between federal centralization and state intervention, now entering a legal phase. The Trump administration has signed an executive order preempting state AI laws and centralizing federal oversight [1][2], following through on the direction telegraphed by a December 2025 order aimed at eliminating state law obstruction of national AI policy [3]. The administration's position — framed as providing regulatory uniformity and preserving US competitiveness against China [4][5] — is simultaneously drawing fire from the left (for creating a governance vacuum) and from libertarian-leaning conservatives who see three specific structural problems with the preemption approach [6].

California has responded with layered executive action. Governor Newsom's formally signed AI workforce executive order [7][8] directs state agencies to study severance pay, subsidized employment, and other mechanisms for workers displaced by AI — treating job displacement as a public policy responsibility rather than a private corporate matter. A separate California executive order, N-5-26, establishes AI certification standards for state procurement [9], creating an accountability framework for AI systems used in state government. Both orders exist in legal tension with the federal preemption claim, and their survival likely depends on whether courts treat them as labor or procurement policy (traditionally state domain) or as AI-specific regulation subject to federal override.

The legal and policy community has mobilized extensively around the preemption question. Law firms including Seyfarth, Latham & Watkins, Baker Botts, and Manatt have published detailed analyses of the federal-state conflict [1][10][11][12]. Manatt argues that the promise of 'federal AI clarity' is largely illusory — centralization in name without enforceable federal standards merely removes state accountability without replacing it [12]. Constitutional analysts at EBG Law examine the limits of executive preemption authority in the absence of congressional legislation [13], while Darrow Everett describes the EO as opening the door to federal-state legal battles [14]. The National Taxpayers Union has raised three specific concerns about the preemption strategy, adding a traditionally conservative voice to the chorus of criticism [6].

The broader governance debate continues to be shaped by the US-China competitiveness frame that the Trump administration has adopted as its primary justification for federal inaction on safety [4][15]. That frame treats AI regulation as a handicap in an international race, while California's workforce and certification orders implicitly treat the same AI progress as an economic shock requiring compensatory policy. Both frames can be simultaneously true — AI can accelerate US economic output while displacing workers — but they point toward incompatible regulatory conclusions. The resilience-based alternative to both restriction and deregulation, which emphasizes biosecurity screening, red-teaming, and infrastructure hardening [16], requires coordinated cross-agency implementation that no current executive order appears to be initiating.