The White House published a National Policy Framework for AI in March 2026 pushing Congress to preempt state AI laws, but as of August 2026 that preemption has not happened — the Senate voted 99-1 to strip a 10-year state-preemption provision from a major bill in mid-2026, and states had already enacted 109 AI laws and 28 data-center laws by July 1. Companies operating in California, Colorado, Illinois, New York and Utah still need full state-by-state AI compliance programs; a unified federal standard is not a near-term planning assumption.
Every few months since late 2025, a new federal proposal has promised to simplify AI compliance into a single national standard. None has passed. For companies deploying AI in hiring, lending, healthcare or consumer-facing products, the practical reality in August 2026 is the opposite of simplification: more states have active AI-specific statutes than at any point before, and the federal push to override them keeps stalling in Congress even as the White House keeps escalating its rhetoric.
What is the White House’s National Policy Framework for AI?
Released March 20, 2026, the framework is a set of legislative recommendations urging Congress to establish a unified federal AI standard and preempt state laws that impose what the administration calls “undue burdens” on AI development and deployment.
The framework builds on Executive Order 14365, issued in December 2025, and covers child safety, consumer protection, data-center energy costs, national security, intellectual property, free speech and workforce development. Notably, it frames itself as “respecting federalism and State rights” while simultaneously calling for federal override of state law — a tension that has shaped nearly every subsequent legislative fight over the framework’s recommendations.
Why did the Senate strip the 10-year state-AI-preemption provision by a 99-1 vote?
Congress considered inserting a 10-year moratorium on state AI regulation into the “One Big Beautiful Bill Act,” but after significant media scrutiny and bipartisan opposition, the Senate voted 99-1 to remove the provision before the bill passed.
That near-unanimous vote is the clearest signal yet that broad, blunt federal preemption lacks political support, even among lawmakers who favor a lighter-touch federal AI standard in principle. State attorneys general, child-safety advocates and some industry groups all opposed the provision for different reasons, which is the coalition dynamic that has repeatedly sunk similar preemption language throughout 2026.
How many state AI laws are already in force in 2026?
As of July 1, 2026, states had enacted 109 AI-specific laws and 28 data-center laws, and AI regulation is described by legal trackers as no longer an emerging frontier but a mainstream compliance obligation.
That volume means a company operating in more than a handful of states cannot realistically wait for federal clarity before building an AI compliance function. The relevant question for most legal and compliance teams is no longer “will there be a state law that applies to us” but “which of the 100-plus already apply, and which of our AI systems trigger them.”
Which states have the most active AI regulatory regimes?
California, Colorado, Utah, Illinois and New York remain the most active jurisdictions, with statutes spanning consumer transparency, healthcare AI use, employment decisions and frontier-model governance.
California’s frontier-model law, SB 53, has become a particular flashpoint because it directly conflicts with federal preemption proposals discussed below. Colorado’s AI Act focuses on algorithmic discrimination in consequential decisions such as hiring and lending. Illinois and New York have concentrated on employment-decision transparency and consumer-facing disclosure requirements, extending frameworks that predate the current federal debate.
What is the Great American AI Act and could it change the picture?
Introduced June 4, 2026 by Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA), the Great American AI Act would nationalize the frontier-model governance approach several states have already adopted, rather than preempting state law outright.
Its bipartisan sponsorship is notable given how partisan most 2026 AI legislation has been, and it represents a materially different strategy than blanket preemption: harmonizing federal rules around the state approach already in place, instead of overriding it. As of early August 2026 the bill remains in committee, but companies tracking the legislative landscape should watch it as the more procedurally realistic path compared to full preemption bills.
How does the FRONTIER Act conflict with California’s SB 53?
The draft FRONTIER Act explicitly preempts state laws covering defined “Covered Subject Areas” of frontier AI, a scope that directly conflicts with California’s SB 53 frontier-model safety and transparency requirements.
This is the sharpest current example of the federal-state collision: SB 53 is already in force and imposes disclosure and safety-testing obligations on frontier-model developers, while the FRONTIER Act draft would nullify exactly that kind of state requirement if enacted. Until one side wins, companies training or deploying frontier models with any California nexus should assume SB 53 governs.
Is federal preemption of state AI laws likely in 2026?
Preemption is unlikely in the near term. Congress has not passed comprehensive federal AI legislation despite multiple bills introduced throughout 2026, and the 99-1 Senate vote against the broadest preemption attempt suggests limited appetite for overriding state law wholesale.
The more probable near-term path is incremental and narrower: harmonizing bills like the Great American AI Act that work with existing state frameworks rather than erasing them, alongside continued case-by-case litigation over specific conflicts like the SB 53–FRONTIER Act dispute.
What does the patchwork mean for AI vendors and enterprise buyers?
AI vendors selling into multiple states now need to support customer-specific configuration for disclosure, logging and human-review requirements, because the strictest applicable state law effectively sets the product baseline for any customer operating there.
Enterprise buyers are responding by pushing more compliance obligation into vendor contracts — requiring AI vendors to warrant that a given model or feature meets Colorado AI Act risk-assessment standards, or that a consumer-facing chatbot satisfies Illinois and California disclosure rules, rather than the buyer building that verification in-house. That shift is showing up in procurement questionnaires and vendor security reviews across regulated industries, and it means technology and procurement teams increasingly need to coordinate on AI vendor due diligence rather than treating it as a purely technical evaluation.
How should companies build compliance around a moving federal-state target?
Build the compliance program around currently enacted state law first, then layer in federal-framework monitoring as a secondary, non-blocking workstream, rather than waiting for federal clarity before acting.
In practice that means: maintaining a live inventory of AI systems mapped against the states where they operate or have effects; prioritizing compliance work in California, Colorado, Illinois, New York and Utah given their regulatory density; assigning legal ownership for tracking both state legislative sessions and the federal preemption fight, since a single team missing either side of the picture is the most common gap; and building AI governance documentation (model cards, risk assessments, disclosure logs) to a standard that satisfies the strictest applicable state law, which tends to also satisfy lighter state and any eventual federal requirement.
Frequently Asked Questions
Has Congress passed a law preempting state AI regulations?
No. As of August 2026, no federal statute preempts state AI laws. A 10-year preemption provision was stripped from a major bill by a 99-1 Senate vote in mid-2026, and other preemption bills remain in committee.
How many state AI laws exist as of mid-2026?
States had enacted 109 AI-specific laws and 28 data-center laws as of July 1, 2026, according to legislative trackers covering the 2026 sessions.
Does California’s SB 53 still apply to frontier-model developers?
Yes. SB 53 remains in force. The draft FRONTIER Act would preempt laws like it if enacted, but as of August 2026 it has not passed, so SB 53 obligations still apply to companies with a California nexus.
What is the Great American AI Act?
A bipartisan bill introduced June 4, 2026 by Reps. Jay Obernolte and Lori Trahan that would nationalize the frontier-model governance approach already used by several states, rather than preempting state law outright.
Should companies wait for federal AI rules before building compliance programs?
No. Given that 109-plus state AI laws are already enforceable and federal preemption has repeatedly failed to pass, companies should build compliance around current state law now and treat federal developments as a secondary monitoring track.
Last updated: August 5, 2026. This article summarizes publicly available legislative-tracking and legal-alert sources and does not constitute legal advice; companies should confirm applicable obligations with AI governance counsel in each operating state.
Written by the Kurums Technology Desk, covering AI strategy, governance and regulatory compliance for technology leaders. Related reading: EU AI Act’s August 2026 Deadline: What Changed for Business Leaders, AI Agents in Corporate Finance: How CFOs Are Deploying Autonomous Systems in 2026, and the Technology Department Hub for more AI governance guides.
Discover more from Kurums | Business Intelligence
Subscribe to get the latest posts sent to your email.
