Federal AI Preemption: What Executive Order 14365 Actually Does
Executive Order 14365 creates a federal strategy for challenging selected state AI laws, evaluating state measures, attaching conditions to some funding, and developing federal reporting and legislative proposals. It does not erase state law by itself. Compliance teams need a versioned obligation register that separates enacted duties from litigation, agency proceedings, funding actions, and proposed legislation.
Federal AI preemption became a concrete executive policy on December 11, 2025, when President Trump signed Executive Order 14365. The order directs federal officials to challenge selected state AI laws, evaluate state measures, consider funding conditions, explore federal disclosure standards, and prepare a legislative recommendation. It does not declare every state AI law void.
That distinction should govern the compliance response. A company must track what law applies today while legal challenges and agency proceedings develop. Deleting a state requirement from the control register because the order criticizes it would confuse a federal strategy with a final legal result.
TL;DR
- Executive Order 14365 establishes a federal policy and directs litigation and evaluation work. It also directs funding and agency work, along with a legislative recommendation. It is not a blanket repeal of state AI laws.
- State duties remain in the compliance register unless controlling law, a court, or another legally effective action changes their application.
- Track each state requirement beside challenges and agency actions. Include funding conditions and effective dates alongside counsel's current interpretation.
- Runtime AI controls can preserve evidence for routed requests, but they cannot decide preemption, suspend statutes, or replace legal monitoring.
Federal AI preemption is a program of actions
The executive order states a policy of sustaining United States AI leadership through a minimally burdensome national framework. It then assigns several different mechanisms.
Section 3 tells the Attorney General to establish an AI Litigation Task Force to challenge state AI laws considered inconsistent with that policy. The listed theories include interstate-commerce limits, preemption by existing federal regulation, and other grounds the Attorney General considers appropriate. A litigation task force can bring arguments to a court. Its creation does not itself produce a judgment against a state law.
Section 4 directs the Secretary of Commerce to publish an evaluation of existing state AI laws. Section 5 addresses conditions for remaining non-deployment funds under the Broadband Equity Access and Deployment program, to the maximum extent allowed by federal law, and tells agencies to assess whether discretionary grants could carry related conditions. These are distinct funding and evaluation actions, not one universal preemption switch.
The White House fact sheet summarizes the order as protection against an inconsistent state compliance regime. It describes litigation, Commerce evaluation, BEAD funding, possible discretionary-grant conditions, and actions by the Federal Communications Commission and Federal Trade Commission. The summary is useful for policy intent, while the signed order controls the assigned steps and qualifications.
The order does not repeal state AI laws
Section 6 directs the FCC chair to initiate a proceeding on whether to adopt a federal reporting and disclosure standard for AI models that preempts conflicting state laws. Section 7 directs the FTC chair to issue a policy statement about circumstances in which the FTC Act could preempt state laws requiring changes to truthful model outputs. Both provisions require later agency work.
Section 8 calls for a legislative recommendation establishing a uniform federal AI policy framework that would preempt conflicting state laws. A recommendation is not legislation. The same section says the recommendation should not propose preempting otherwise lawful state measures on child safety, state government procurement and use of AI, AI compute and data-center infrastructure except certain permitting reforms, and other topics later determined.
Section 9 also says implementation must remain consistent with applicable law and that the order creates no enforceable right or benefit against the United States. These clauses reinforce the operational point: scope and legal effect will depend on later actions and their authority.
My view is blunt. Any control team removing state-law obligations solely because this order exists is making a category error. The prudent register keeps the enacted duty visible and adds the federal challenge, proceeding, or funding action as a separate legal event.
Build a state-law change register around legal effect
Put a large state map on the review wall, but do not let colored pins become the system of record. The useful artifact is a versioned register with one row per obligation and a dated interpretation.
For each state measure, retain the jurisdiction, citation, covered role, regulated activity, effective date, required control, evidence owner, enforcement authority, and counsel's conclusion. Add fields for a federal challenge, court, docket, agency proceeding, funding condition, injunction, final judgment, appeal, and next review date.
The Colorado AI Act compliance guide and Texas TRAIGA checklist illustrate why state analysis remains granular. Different statutes use different role definitions, triggers, exclusions, enforcement models, and effective dates. A national policy dispute does not flatten those differences overnight.
Use three legal-effect states rather than a binary active or repealed label:
- Operative: the organization treats the requirement as applicable and tests its controls.
- Contested: a challenge or proceeding may change application, while the current operating position remains documented.
- Displaced or unenforceable: counsel cites the controlling statute, rule, order, or final judicial action producing that result and states its scope.
The register should also retain historical versions. An investigation in 2027 may ask what the organization reasonably understood on a particular 2026 date, before a later ruling changed the analysis.
Keep legal interpretation separate from runtime evidence
A public-sector or regulated enterprise may already map federal, state, procurement, and framework duties through a broader public-sector AI compliance program. Executive Order 14365 adds monitoring work for legal and government-affairs owners. It does not convert every state obligation into a gateway rule.
Runtime controls support narrower questions. They can show which authenticated principal sent a routed request, which destination received it, what data classification applied, which policy version ran, and whether the action was permitted, redacted, or denied. That evidence may help prove operation of a control required by a still-applicable state law.
Legal owns applicability and preemption analysis. Government affairs follows federal and state developments. Compliance translates the current interpretation into obligations. Engineering implements testable controls. Internal audit samples evidence against the dated register. Keeping those roles separate prevents an infrastructure team from silently deciding a constitutional question through a configuration change.
DeepInspect
DeepInspect can enforce policies and create decision records for deliberately routed, authenticated HTTP traffic to AI endpoints. It can help demonstrate that a current state-law control operated on those requests. It cannot determine that federal law preempts a state measure, monitor every legal proceeding, govern traffic that bypasses the route, or replace counsel's dated applicability decision.
Book a demo today.
Frequently asked questions
- Did Executive Order 14365 invalidate every state AI law?
No. It establishes federal policy and directs litigation, evaluation, funding, agency, and legislative actions. Whether a particular state provision is preempted or otherwise unenforceable depends on the legal authority and outcome applicable to that provision.
- Should companies pause state AI compliance projects?
Not solely because of the order. Keep implementing obligations that counsel considers operative. Record any challenge or expected agency action and use material legal changes as review triggers rather than assumptions.
- What does the AI Litigation Task Force do?
The order assigns it responsibility for challenging state AI laws considered inconsistent with the federal policy, including through constitutional and existing-preemption arguments. Courts still determine the cases presented to them.
- Are all state-law topics targeted equally?
No. Section 8's legislative instruction identifies areas the recommendation should not propose preempting when state laws are otherwise lawful, including child safety, state procurement and use, and specified infrastructure matters. Later agency and judicial actions may have their own narrower scopes.
- What evidence should an enterprise retain now?
Retain the obligation register, source text, dated legal interpretation, change approvals, implemented control tests, and production samples. When the interpretation changes, preserve the old version and record who approved the new operating position.