What happened

A December 2025 White House executive order framed state AI laws as a national-security and competitiveness problem, arguing that “State-by-State regulation” creates a “patchwork” that is especially painful for startups (White House). It singled out Colorado’s “algorithmic discrimination” language as an example of state rules that could force model outputs toward what the order calls “false results” to avoid differential impact (White House).

In March 2026, the White House published a national “legislative framework” for AI that calls for Congress to preempt state AI laws that impose “undue burdens,” aiming for “a minimally burdensome national standard… not fifty discordant ones” (White House framework). At the same time, the framework insists it’s not proposing a shiny new federal AI rulemaking body — and it carves out room for states to keep enforcing traditional police powers (fraud, consumer protection, child safety), zoning, and their own procurement rules (White House framework).

Lawfare’s read is that this isn’t classic preemption at all. It’s an executive-branch “managed federalism” approach: keep state authority technically alive while steering it with coordination, incentives, and the implied threat of federal challenge (Lawfare).

The non-obvious angle: “pre-emption without Congress” is the real product

Everyone loves arguing about whether AI should be regulated. The higher-signal question is: who gets to regulate it — and how fast they can make everyone else stop.

The White House position is basically: AI development is interstate, therefore state-level developer rules are illegitimate friction; but state-level enforcement on harms and consumer protection is fine; and state procurement can stay a lever (White House framework). That’s a political compromise disguised as a constitutional principle. It keeps “states’ rights” alive, but it pushes the high-value control surface (developer obligations) up to the federal level.

The mechanism test (who pays, who audits, who carries liability)

Who pays: The executive order language repeatedly ties federal action to stopping “onerous” state regulation — the kind of framing that’s designed to justify conditional funding or agency pressure, even before Congress writes a single line of statutory preemption (White House; Lawfare). If the White House can make state AI rules costly through federal levers, the argument about formal doctrine becomes… decorative.

Who audits: The framework says “Congress should not create any new federal rulemaking body to regulate AI” (White House framework). That sounds deregulatory, but it also dodges the staffing problem. If you preempt states and don’t build a federal audit capacity, oversight doesn’t vanish — it relocates into procurement checklists, lawsuits, and whichever agency already had jurisdiction over “something adjacent.”

Who carries liability: The framework is explicit that states “should not be permitted to penalize AI developers for a third party’s unlawful conduct involving their models” (White House framework). That’s a liability shield posture. The pressure will move downstream: onto deployers, integrators, and buyers — i.e., the people who can’t afford a Supreme Court case but can afford a compliance vendor.

So what’s actually being centralized?

• Developer rules: States are told not to regulate development. That’s the highest-leverage point for shaping the market.
• Baseline definitions: A federal standard means one set of definitions becomes the template everyone builds against — whether it’s good or not.
• Enforcement posture: Even before legislation, executive-branch coordination can chill state experimentation. The silence is the point.