A December 2025 executive order designed to preempt state AI laws is now in active implementation, with an FTC deadline of March 11 that could weaponise federal consumer protection law against state-level algorithmic accountability requirements.
While the AI industry watches benchmark scores and model releases, a quieter transformation is underway in American governance. The Trump administration's executive order on AI — signed in December 2025 — is systematically dismantling the patchwork of state AI protections that have emerged over the past three years. Some of those protections were imperfect. But the replacement isn't a better framework; it may be no framework at all.
What Happened
On December 11, 2025, the White House issued an executive order titled "Ensuring a National Policy Framework for Artificial Intelligence." The order establishes an AI Litigation Task Force within the Department of Justice to challenge state AI laws deemed inconsistent with federal AI policy. It directs the FTC to issue a policy statement by March 11, 2026, classifying state-mandated bias mitigation requirements as per se deceptive trade practices — a legal manoeuvre that would effectively turn the federal consumer protection framework against state consumer protection laws. The order also creates mechanisms to condition federal funding on states' compliance with the emerging national framework.
This federal push is happening against a backdrop of state-level retreat. Colorado, which passed what was widely regarded as the US's most comprehensive high-risk AI bill (SB 24-205) in 2024, has delayed implementation to June 30, 2026 (from February 1, 2026) and is actively reconsidering the law through a special legislative session. Industry advocates successfully pressed for narrower definitions and longer timelines. California's SB 53 — a transparency law requiring major AI companies to publish safety details and protecting AI whistleblowers — was positioned as a lighter-touch alternative to the vetoed SB 1047, carefully calibrated to avoid triggering federal preemption. The One Big Beautiful Bill Act had included a 10-year moratorium on new state AI regulations; the Senate stripped it out on bipartisan grounds, forcing the administration to pursue the same goals through executive action.
Why It Matters
The practical effect of the executive order is to transform the US regulatory landscape from a complex patchwork of state protections into something approaching a federal standard explicitly designed to be "minimally burdensome." The EO's stated goals include discouraging state authority through litigation and spending, targeting state rules on algorithmic transparency and bias mitigation, and promoting national AI standards that prioritise innovation over precaution.
The FTC deadline is the most immediate pressure point. If the FTC follows the EO's direction and classifies state bias-mitigation requirements as deceptive trade practices, it will create a federal legal basis for AI companies to challenge any state consumer protection law that requires algorithmic impact assessments. This would be the most aggressive federal preemption of state consumer law since the financial crisis-era federal banking regulator arguments. Specifically: Colorado's SB 24-205 required developers of high-risk AI systems — those making consequential decisions in employment, credit, housing, and healthcare — to conduct impact assessments and disclose AI involvement to affected consumers. These are not radical requirements. The EU AI Act contains comparable provisions. But under the EO framework, they become targets.
The specific legal mechanism — using the FTC's deceptive trade practices authority — is particularly aggressive because it does not merely preempt state regulation; it frames state regulation as itself harmful to consumers. If that argument succeeds, states that enforce algorithmic accountability requirements could face federal legal challenge not just on preemption grounds but on the grounds that enforcement itself constitutes a deceptive practice.
Wider Context
The context here is a widening divergence between US and EU approaches. The EU AI Act is in active implementation: prohibitions on unacceptable-risk AI have been in effect since August 2024, with obligations for high-risk systems taking effect through 2026 and 2027. The US, by contrast, is moving toward a minimalist federal standard explicitly oriented around competitive advantage and innovation speed.
This divergence will create increasingly significant compliance asymmetries for multinational companies. A US AI company operating in Europe is subject to EU AI Act requirements; operating in the US, it is now potentially subject to federal legal action if it complies with certain state laws. For companies building systems that make consequential decisions — hiring algorithms, credit scoring, medical triage — the question of which regulatory framework governs will become a live legal question, not merely a compliance exercise.
The Trump administration's framing is that fragmented state regulation creates "regulatory fragmentation" that impedes innovation and disadvantages US companies relative to Chinese competitors. That argument has genuine force in some domains. But it elides a distinction: the EU's approach is also a form of federal preemption — a single regulatory framework replacing member state variations. The difference is that the EU framework contains substantive requirements; the emerging US framework appears to be substantively empty. The national AI standard that preempts state laws is a standard of minimal interference, not robust protection.
The Singularity Soup Take
The US is not accidentally becoming the world's most permissive major AI jurisdiction — it is doing so by design. The EO is a deliberate choice to prioritise competitive advantage and innovation speed over the regulatory precautions that other democratic jurisdictions treat as baseline requirements.
The argument for this approach is that AI is a competitive domain where regulatory friction costs American companies market position. That has some force. The counter-argument is that the risks being mitigated by bias assessment requirements and algorithmic transparency are real and already documented — and that removing those requirements primarily advantages the companies deploying AI systems, not the workers and consumers affected by their decisions.
The FTC deadline of March 11 deserves to be treated as a moment of genuine significance. If the FTC broadly frames its policy statement, it will have resolved a fundamental tension in AI governance in favour of developer convenience over consumer protection — and that resolution will be very difficult to reverse. State legislatures that have spent years building AI accountability frameworks will face federal legal risk for enforcing them. The companies those frameworks were designed to constrain will be among the largest beneficiaries. That is not a neutral policy outcome, and it should not be reported as one.
What to Watch
The March 11 FTC statement is the most immediate signal. If it is narrowly scoped and limited to specific types of bias requirements, the EO's practical reach may be limited; if it is broadly framed to cover any state algorithmic accountability rule, the implications are sweeping. Watch how California responds — SB 53 was carefully designed to avoid preemption triggers, but a broad FTC statement may require the state to recalibrate. And watch the EU-US AI dialogue: if European regulators signal that the US's minimalist approach threatens data-sharing or mutual recognition arrangements, that could create counterpressure on the administration from the tech industry itself, which has significant EU exposure and does not want a regulatory war on two fronts.
Sources
Paul Hastings — President Trump Signs Executive Order Challenging State AI Laws
Latham & Watkins — AI Executive Order Targets State Laws and Seeks Uniform Federal Standards
The Verge — Meet the new tech laws of 2026
University of Denver — Colorado is pumping the brakes on first-of-its-kind AI regulation
The White House — Ensuring a National Policy Framework for Artificial Intelligence