Version 4

The Trump administration has signed an executive order that law firms variously describe as 'preempting' or 'challenging' state AI laws [1][2][11], while simultaneously declining to impose substantive federal AI safety regulations — a combination creating what analysts call a governance vacuum. California has responded with two executive orders: a workforce displacement order directing state agencies to study severance pay and other support mechanisms for AI-displaced workers [20][21], and Executive Order N-5-26 establishing AI certification and procurement standards for state government [24][17]. The White House has also released a formal National Policy Framework for AI document [7], but an accompanying AI security executive order was scrapped hours before signing [4][5]. An expanding legal analysis ecosystem — now spanning more than a dozen major law firms — is parsing the constitutional and practical limits of the federal preemption claim.

The structural problem is that federal preemption removes state-level accountability without replacing it with enforceable federal standards, leaving workers, consumers, and state agencies exposed to AI's economic and safety impacts with no clear regulatory backstop. California's two-track response — workforce policy and procurement standards — is the most concrete attempt by any government actor to fill that vacuum, making the legal survival of those orders a test case for whether any government can govern AI in the near term.

American AI governance in 2026 is defined by a structural conflict between federal preemption and state intervention, now entering a phase of intensive legal analysis and contested framing. The Trump administration has signed an executive order aimed at preempting state AI laws and centralizing federal oversight [1][2], following a December 2025 order that explicitly targeted state law obstruction of national AI policy [3]. Simultaneously, the administration scrapped a separate AI security executive order hours before signing, with the stated rationale that regulation would impede US AI leadership relative to China [4][5][6]. The White House has published a formal National Policy Framework document [7] as a companion to these actions, though analysts question whether it establishes enforceable standards or merely articulates aspirational goals.

A notable legal framing distinction has emerged: while most law firms analyzing the preemption EO use the language of 'preemption' [1][8][9][10], Paul Hastings describes the order as 'challenging' state AI laws rather than definitively overriding them [11]. This framing difference may reflect genuine ambiguity in the EO's legal mechanism — executive orders cannot unilaterally preempt state laws without congressional authorization in the same way statutes can, and constitutional analysts have flagged this gap [12][13]. The legal community's response to the EO has expanded substantially, with Cleary Gottlieb [14], Wiley Law [15], Morgan Lewis [16], Ropes Gray [17], and Alston Privacy [18] joining earlier analyses from Seyfarth, Latham & Watkins, Baker Botts, and Manatt. The National Taxpayers Union, representing a conservative voice, has raised structural concerns about the preemption strategy [19].

California has built a layered governance response at both the executive and legislative levels. Governor Newsom's formally signed AI workforce executive order [20][21][22][23] directs state agencies to study severance pay, subsidized employment, and other mechanisms for workers displaced by AI — framing job displacement as a public policy obligation rather than a private corporate matter. Executive Order N-5-26 [24][17][25] establishes AI certification standards and procurement requirements for state government contracts, requiring vendors to meet trust and safety standards [15]. Better Markets, a progressive financial reform organization, has endorsed the workforce order as 'a good start' [26], while academic observers are beginning to assess the data-driven framework embedded in the order [22]. California is also pursuing parallel legislative action: SB951, currently in session [27], suggests the state is building statutory foundations alongside the executive order architecture.

The governance competition between federal preemption and state intervention is underpinned by incompatible framings of AI's economic impact. The Trump administration treats AI regulation as a competitive handicap in a race against China [6][28][29], implying that AI's net effect is positive and that constraints are the problem. California's workforce and procurement orders treat the same AI progress as producing economic shocks requiring compensatory policy — severance, retraining support, and vendor accountability — implying that AI's distributional impacts require active government response. Both framings can be simultaneously true, but they lead to incompatible regulatory conclusions. The resilience-based alternative advanced by analysts including Sayash Kapoor — emphasizing biosecurity screening, red-teaming, and infrastructure hardening rather than access controls or preemption [30][31][32] — remains a minority position with no current executive order implementing it.