AI Policy Wiki
Dashboard
⚲
AI Policy Wiki › Concepts & Frameworks

AI Federalism

medium confidence · updated 2026-09-26

The contest over how regulatory authority over AI is allocated between the US federal government and the states — the preemption fight, the state-preservation coalitions resisting it, and the resulting de facto patchwork.

Contents

AI federalism refers to the contest over how regulatory authority over AI is allocated between the US federal government and the states: which level of government sets the binding rules for AI, and how that authority is contested and preempted. The term covers the federal preemption effort, the state-preservation coalitions resisting it, and the de facto regulatory patchwork that has emerged in the absence of comprehensive federal frontier-AI law.

The topic is narrower than the catalog of individual state laws in State-Level AI Regulation, distinct from OpenAI's May 2026 framing of states converging into a de facto national standard in Reverse Federalism, and distinct from Tim Wu's academic framework on center–local–market interplay in the US and China in Techno-Federalism. Where those address, respectively, the catalog, one named coinage, and one scholarly model, AI federalism concerns the political-legal contest over who governs.

Structural situation (mid-2026)

As of May 2026, the United States has no comprehensive federal frontier-AI statute. The operative regime is therefore set, in practice, by the states, the dynamic that State-Level AI Regulation describes as "the principal field of US AI regulation." Three trifecta states reached finish lines on overlapping but distinct architectures within a single 48-hour window in late May 2026 (Source: iapp.org):

  • Illinois SB 315 — the first state mandate for annual independent third-party audits of frontier-AI safety practices.
  • Connecticut SB 5 — a bundle of frontier-model risk-reporting, ADMT-disclosure, and AI-companion regimes.
  • New York Safe By Design Act — children's online-safety protections folded into the FY27 budget.

Together with California's SB 53 and the New York RAISE Act, they constitute the emerging patchwork.

Federal preemption pressure

The Trump administration has sought to preempt state-level AI regulation in favor of a single, lighter federal touch, most concretely via the December 2025 draft executive order on state-AI preemption, alongside congressional efforts including the proposed federal AI moratorium on state enforcement. White House AI advisor David Sacks has been the most prominent voice against a heavy regulatory hand. The argument advanced on the preemption side holds that a fifty-state patchwork raises compliance costs, fragments the market, and cedes ground in the US–China race. Industry coalitions including the Chamber of Progress, CCIA, and NetChoice generally align with the preemption side on cost-and-fragmentation grounds.

A July 18, 2026 analysis by The Information concluded that the administration's light-touch AI agenda — set by the January 2025 executive order revoking Biden-era AI efforts — had been upended: hundreds of states, cities, and counties passed data-center moratoriums, and the administration itself shifted toward interventionist measures including restrictions on top AI models in the U.S. (Source: theinformation.com). See Data Center Siting / AI Power Politics, Export Controls (AI).

Critics of the preemption campaign have argued that it is not limited to displacing state law. A September 2026 Center for American Progress report by Nicole Alvarez and Adam Conner states that preemption supporters argue Congress needs time to develop a national framework, "despite state laws currently being the main source of frontier AI oversight," and claim that existing federal law and voluntary company standards suffice; it argues that Sen. Ted Cruz's September 2025 legislative framework and Executive Order 14365 show the approach "is not truly deferential to existing federal law," pointing to Cruz's SANDBOX Act, which would let the OSTP director waive or modify federal rules for AI deployment. "The problem, in the eyes of the act's proponents," the authors write, "is not just the inconsistency of state laws, but the existence of federal rules" (The SANDBOX Act Uses AI Policy as a Trojan Horse for Unfettered Deregulation (Center for American Progress, September 2026)).

Intra-party divisions surfaced again in July 2026. Sen. Marsha Blackburn's Oval Office pitch to President Trump pairing her children's-online-safety package with a narrow preemption of state AI laws — backed by David Sacks and Kevin Hassett — stalled after Sen. Ted Cruz objected and the White House pulled a prepared endorsement post, per July 21, 2026 reporting (Source: washingtonpost.com).

A Senate vehicle for a federal standard emerged at the end of July 2026. At a July 30, 2026 Senate Commerce telecommunications subcommittee hearing, Sen. Amy Klobuchar (D-MN) said she and Majority Leader John Thune (R-SD) aimed to produce a bill setting federal AI standards by the end of the August recess, telling the panel, "We can't just stand down and not have coordinated standards" on frontier models (Source: insideaipolicy.com). The hearing also covered telecommunications infrastructure buildout, cybersecurity, and regulatory oversight of frontier models. Ranking member Sen. Maria Cantwell (D-WA) used the same hearing to press for mandatory guardrails on frontier models and to reject industry self-certification, citing cyber and biosecurity risks and pointing to the OpenAI disclosure of the previous week (Source: insideaipolicy.com). Neither the bill's text nor its position on preemption of state law had been made public.

OpenAI had set out the fullest version of its preemption position two months earlier. Its June 2, 2026 federal blueprint asks Congress, once a comprehensive federal framework is in place, to "preempt state laws that seek to regulate the same frontier safety risks, creating a single national framework that combines strong safeguards with regulatory certainty." The ask is bounded rather than general: three areas are carved out in which states "should continue serving as laboratories of democracy" — youth protection, electricity and environmental policy, and AI education and literacy. The document also frames the three state frontier statutes as "the foundation for federal frontier safety legislation—not its endpoint," positioning preemption as the terminus of a sequence in which state requirements are first federalized and then extended (Democratic Governance of Frontier AI: A blueprint for a federal framework (OpenAI, June 2026)).

Frontier developers took public positions on the same question in early August 2026. OpenAI argued on August 3 that "America ultimately needs national AI standards established through Congress," on the grounds that a patchwork of state requirements could slow American developers while the People's Republic of China pursues a national strategy, and that divergent state approaches to frontier-AI safety would produce "confusion about which models can be deployed, when, and to whom." Its fallback position, should Congress not act quickly, is that states converge on common approaches rather than diverge (Keeping America out in front on AI (OpenAI Global Affairs, August 2026)). Anthropic told the Senate offices working on the frontier bill that it wants the testing regime strengthened and holds concerns about accompanying language preempting state AI laws (Source: washingtonpost.com).

State-preservation resistance

A bipartisan state-preservation coalition has formed in opposition. When New York enacted the Safe By Design Act, AG Letitia James led 43 state attorneys general in opposing the federal KIDS Act on preemption grounds the same day, with states enacting and defending their own authority simultaneously (Source: iapp.org). The resistance is not purely partisan: Republican state-level actors have advanced Tenth-Amendment-grounded pushback against federal preemption, including the posture behind Missouri SB 1012 — Artificial Intelligence Non-Sentience and Responsibility Act and the House-side state-preservation stance of John Moolenaar (Source: Missouri SB 1012 — Artificial Intelligence Non-Sentience and Responsibility Act). On the House side, the SCALE Act reflects a Republican variant of the federal-role debate.

Advocacy groups pressed the state-preservation position on the Senate in September 2026. A letter of September 16, 2026 to Majority Leader John Thune and Minority Leader Chuck Schumer, led by Americans for Responsible Innovation and signed by 20 organizations and five individuals, including New York Assembly member Alex Bores, lead Assembly sponsor of the RAISE Act, sets a federal floor as one of three minimum requirements for any Senate framework: Congress should "establish a strong floor with regard to AI regulations while preserving the critical rights of the states to establish additional regulations beyond those set by Congress," and "should not eliminate strong state laws in exchange for weaker federal standards." The letter names no bill (Senate coalition letter on redlines for AI safety legislation (Americans for Responsible Innovation, September 2026)).

State enforcers made the same demand in a request for federal regulation. A letter dated September 23, 2026 from the attorneys general of 24 states, the District of Columbia and American Samoa, led by New York's Letitia James and New Jersey's Jennifer Davenport and addressed to Speaker Mike Johnson, Majority Leader John Thune and Minority Leaders Hakeem Jeffries and Chuck Schumer, asks Congress "to immediately establish comprehensive federal regulation and safety protocols for frontier artificial intelligence," including mandatory federal oversight of safety testing and government-led incident response. Its last request is "explicit preservation of state authority by prohibiting the preemption of state laws and granting state officials full authority to enforce these federal protections." The signers state that they are already acting under state law, citing California SB 53, New York's RAISE Act and Illinois SB 315 (Letter from state attorneys general to congressional leadership urging federal frontier AI regulation (September 2026)). Both the letter and OpenAI's June 2026 blueprint ask Congress for a federal frontier framework; they differ on preemption, which the blueprint asks Congress to apply to state laws regulating the same frontier safety risks (Democratic Governance of Frontier AI: A blueprint for a federal framework (OpenAI, June 2026)).

Relation to policy and the federalism frame

The contest inverts the textbook order in which the federal government sets the framework and states act as "laboratories of democracy" within it. In AI, states are setting the operative compliance regime for nationally deployed products, with federal policy reacting, the dynamic OpenAI labeled reverse federalism. Because a single national market cannot easily comply with one state's rules and ignore another's, the strictest state effectively exports its standard, an application to AI of the California effect. AI federalism therefore concerns not only formal authority but which jurisdiction's substantive choices become the de facto national floor.

Relationships

Provenance note. Built from existing legislation and concept pages plus the IAPP May 28 round-up and Inside AI Policy coverage; confidence: medium.

Categories: Concepts & Frameworks | us | framework | regulation | competition | liability | ai-policy
What links here (64)