Trump's AI Executive Order vs. State Hiring Laws in 2026

No. Trump’s December 2025 executive order does not override state AI hiring laws. Executive Order 14365 directs the U.S. Department of Justice to challenge those laws in court, but an executive order cannot repeal a state statute on its own. NYC Local Law 144, Illinois HB 3773, and California’s FEHA automated-decision rules stay fully enforceable until a court or Congress says otherwise. If you run AI anywhere in your hiring funnel, you still have to comply, today, while the fight plays out.

That is the gap between two headlines that landed in the same three weeks. This is the operator’s version of what to do about it, not legal advice and not a law-review article. If you are a Head of Talent or a founder hiring across several states with no employment counsel on retainer, this is your “what do I build this quarter” plan.

Two headlines, one impossible week

On December 11, 2025, President Trump signed an executive order promising to dismantle the patchwork of state AI laws. Three weeks later, on January 1, 2026, Illinois’s new AI anti-discrimination law took effect anyway. Same employer, opposite signals, same calendar page.

That is the whiplash People teams are living in. One source says the rules are about to disappear; another says a fresh deadline just went live. You cannot pause hiring while Washington and the states argue, and that argument could run for years. So the real question is not “who wins the preemption fight,” it is “what do I build that survives whichever way it breaks?”

The reassuring part: the answer is more stable than the news cycle. Employment lawyers are unanimous on the posture, and the strictest requirements across every live law collapse into the same short list of things to build.

What Executive Order 14365 actually does (and doesn’t)

EO 14365, “Ensuring a National Policy Framework for Artificial Intelligence,” signed December 11, 2025, is an instruction to the federal government to attack state AI laws, not a law that erases them. According to the White House text and analysis from Fisher Phillips, it does four things:

• Stands up a DOJ AI Litigation Task Force within 30 days whose sole job is to sue over state AI laws deemed “inconsistent” with the order, on interstate-commerce, preemption, or “otherwise unlawful” grounds.
• Tasks Commerce with publishing an evaluation of “onerous” state AI laws within 90 days.
• Ties federal broadband (BEAD) funding to states steering clear of such laws.
• Directs the FTC and FCC to take preemption-supporting actions within 90 days.

Here is the part that matters for your hiring process: an executive order cannot invalidate a state law. Preemption requires an act of Congress or a valid federal regulation, and ultimately a court has to agree. Every law firm that weighed in says the same thing. As Fisher Phillips put it, all current and pending state and local AI laws “will remain enforceable unless and until a court blocks them.” White & Case and NPR reached the same conclusion.

Worse for planning purposes: it is not even clear that hiring laws are protected if the federal side wins. The EO’s explicit carve-outs (child safety, AI infrastructure, state government procurement) do not clearly cover employment-related AI laws, which is exactly why workplace rules are a likely battleground. One expert told SHRM they expect workplace rules to survive as carve-outs, but that is a prediction, not a guarantee. You cannot build a hiring process on a guess about how a lawsuit lands.

The state laws that are live right now

Three regimes are in force today and apply the moment you use AI to screen, rank, score, or recommend candidates. Here is what each one actually requires.

NYC Local Law 144

If you use an Automated Employment Decision Tool (AEDT) on candidates in New York City, you must do three things, per the NYC rule:

1. Commission an independent bias audit within the prior 12 months (the auditor cannot have a financial stake in the tool).
2. Publicly post the audit’s adverse-impact results: selection and scoring rates plus impact ratios, measured against the four-fifths (80%) rule.
3. Notify candidates at least 10 business days before the tool is used.

Penalties run $500 to $1,500 per violation, per day. A common excuse is “nobody’s been fined yet.” That window is closing. A December 2025 New York State Comptroller audit found enforcement had been minimal and that the agency committed to fixing it. Lax enforcement is not the same as a safe harbor, and it is tightening.

Illinois HB 3773

Effective January 1, 2026, Illinois amended its Human Rights Act so that using AI with a discriminatory effect in recruiting, hiring, promotion, discipline, or discharge is a civil-rights violation, per the National Law Review. Three things stand out:

• Intent doesn’t matter. Discriminatory effect is enough.
• You cannot use zip codes as a proxy for protected classes.
• There is an affirmative notice obligation whenever AI is used to “influence or facilitate” a covered decision.

It applies to employers with at least one employee for 20 or more weeks in the year, so the small-team exemption many founders assume they have does not exist here.

California FEHA automated-decision systems

California’s FEHA regulations on Automated Decision Systems took effect October 1, 2025. Per Manatt, they cover any tool that “screens, scores, ranks or recommends candidates, even where humans retain final decision-making authority.”

Read that last clause twice. A human in the loop is not automatically a loophole. Whether your anti-bias testing exists, how recent it is, how broad it is, and how you responded to what it found all become central evidence in a discrimination claim. A token human rubber-stamp does not exempt you. A real review gate with a record is what protects you.

Colorado: a cautionary tale in regulatory whiplash

If you ever feel tempted to bet your hiring process on one law’s survival, look at Colorado.

Colorado’s original AI Act (SB 24-205) was enacted in May 2024 to take effect February 1, 2026. Then the timeline came apart: the date was pushed to June 30, 2026, the law was federally enjoined in April 2026, and then it was repealed and replaced by SB 26-189, signed May 14, 2026. The original Act never took effect at all. The replacement, a lighter notice-and-transparency regime, takes effect January 1, 2027, with pre-use notices, adverse-outcome explanations, and meaningful human review, per Norton Rose Fulbright.

Enacted, delayed twice, enjoined, repealed, rewritten, all before it ever applied. That is the whole argument in one statute: betting your process on any single regime surviving is a losing bet. The laws will keep moving. Your hiring system should not have to move with each one.

What employment lawyers actually recommend

Across every firm that published guidance, the advice converges on one phrase: stay the course, and standardize on the strictest common denominator.

Fisher Phillips: “Stay the course on state-law compliance. Colorado, California, Illinois, New York, and other state requirements are still on track unless and until courts say otherwise.” The National Law Review’s “thread the needle” guidance is more concrete: identify the AI laws in every state you operate in, then “consider whether you should adopt one global policy that follows the most restrictive law,” document each system’s intended use and human-oversight plan, periodically test models on real-world scenarios, keep records of mitigation, and “draft with an eye toward flexibility if state laws are challenged or preempted.”

The lawyers are right, and they also stop exactly where the work begins. “Adopt one global policy” and “document everything” are correct, but they are not a system. Somebody has to turn them into the actual primitives in your hiring stack. That is the part nobody writes down.

Build for the strictest common denominator: five durable primitives

Here is the good news the news cycle buries. When you line up the strictest requirements across NYC, Illinois, California, and even repealed-Colorado, they collapse into the same five primitives. Build these once and you are covered whether Washington wins, the states win, or it settles somewhere in between.

This is exactly the layer Kit ships as defaults, so the same hiring stack stays defensible no matter how the preemption fight breaks.

A per-decision audit trail

Every law in the patchwork eventually asks the same question: what happened, and what role did the AI play? If a candidate or regulator asks why someone was rejected, “the system did it” is the wrong answer. A timestamped log of who advanced or rejected whom, at which stage, and where AI sat in that decision is the record Illinois recordkeeping and California’s “response to identified risks” standard both lean on. Kit logs this by default on every pipeline action, so the audit trail exists before you ever need it.

Human-in-the-loop review gates

California already settled the debate: a human’s final authority does not exempt the tool, but a real review gate is your defense. The distinction is a record. A configurable, mandatory human-review step before a consequential outcome is what turns “AI made the decision” into “a person reviewed and decided, here is the proof.” Kit lets you require team review and decision steps before any reject or advance, and it captures who reviewed and when.

Candidate-facing AI disclosure

Illinois, NYC, and Colorado all require telling candidates that AI is in the loop, in plain language, before they are evaluated. The mechanics differ (NYC wants 10 business days of lead time; Illinois wants affirmative notice) but the primitive is one thing: a disclosure surface candidates actually see. Building this into the application flow, rather than burying it in a privacy policy, is the difference between a notice that counts and one that doesn’t.

Bias-audit-ready evaluation records

NYC’s independent bias audit needs your decision data in a specific shape: selection and scoring rates and impact ratios, measured against the four-fifths rule. California’s anti-bias testing wants the same underlying numbers. If your evaluation data is trapped in a screening vendor’s black box, you cannot produce it on demand and you cannot defend it. The fix is keeping decision and scoring data exportable in the format an auditor expects, so an audit is a report you run, not a fire drill.

Per-jurisdiction notice and adverse-action workflows

NYC wants 10-day notice timing; Colorado’s coming regime wants adverse-outcome explanations; Illinois and California each want their own notices. Rather than hard-coding one state’s rules, the durable move is one workflow engine with per-jurisdiction settings, so a candidate in New York gets the NYC timeline and a candidate in Illinois gets the Illinois notice, from the same pipeline. One system, many jurisdictions, no rebuild when the next state passes a law.

The no-regrets posture

Strip away the politics and the litigation, and you are left with a posture that is correct no matter who wins: transparent, auditable, human-reviewed hiring with the receipts attached.

It is compliant if federal preemption wins, because it is just good hiring hygiene. It is compliant if the states win, because it already meets the strictest of them. And it is compliant under any settlement in between, because the five primitives do not depend on which regime survives. You are not betting on a court. You are building the thing that is defensible either way.

That is the whole point. You cannot control whether EO 14365 holds up, whether the DOJ task force wins, or which state passes the next law. You can control whether your hiring process logs its decisions, keeps a human in the loop, tells candidates the truth, and can produce its records on demand. Do that, and the regulatory whiplash stops being your problem.

Kit gives you those primitives as defaults, not as a compliance bolt-on, because transparent and auditable hiring is just better hiring. If you want a pipeline that holds up whichever way the rules break, start a free trial and build on the strictest common denominator from day one.

This article is operational guidance, not legal advice. For your specific obligations, consult employment counsel.