AI Hiring Laws and Compliance: What US Employers Must Do in 2026

The 2026 picture

Two things happened at once. Federal AI-specific guidance got thinner, and state law got denser. The EEOC pulled its AI technical assistance documents from its website in early 2025 after the January 23, 2025 executive order "Removing Barriers to American Leadership in Artificial Intelligence," and an April 23, 2025 executive order, "Restoring Equality of Opportunity and Meritocracy," told agencies to deprioritize disparate-impact theories of liability. None of that changed the statutes. Title VII, the ADEA and the ADA are still on the books, and private plaintiffs can still bring disparate impact claims. Meanwhile New York City, Illinois, California, Colorado and Texas wrote their own rules, so your obligations now depend on where your candidates live.

Is it legal to use AI in hiring?

Yes. No US law bans AI in hiring. The law holds you responsible for the result: if an AI screening tool produces a discriminatory outcome, the employer is on the hook, and "the vendor built it" is not a defense. Several state and city laws also require you to disclose AI use, audit for bias, or both.

The clearest signal of where liability lands is Mobley v. Workday, in the Northern District of California. In May 2025 the court granted preliminary certification of a nationwide ADEA collective covering applicants aged 40 and older rejected through the vendor's screening features, and in July 2025 it expanded that collective to applicants whose applications were scored, sorted, ranked or screened by the vendor's HiredScore AI features. The case is still in litigation, but the practical point is already made: algorithmic screening at scale creates exposure at scale.

Do employers have to tell candidates they are using AI?

In several jurisdictions, yes. New York City requires at least 10 business days of advance notice before an automated employment decision tool is used on a candidate. Illinois requires notice when AI is used in an employment decision, and separately requires notice and consent before AI analyzes a video interview. There is no federal notice duty, but disclosure is becoming the default expectation.

Disclose regardless of jurisdiction. A candidate who learns after the fact that a model ranked them tends to escalate. One told up front, in a plain sentence, usually does not.

The laws that actually apply to hiring, compared

This covers the US laws that specifically regulate AI or automated tools in employment decisions, as of July 2026. At least one of them is likely to move again.

What NYC Local Law 144 costs you if you ignore it

Penalties run up to $500 for a first violation and $500 to $1,500 for each subsequent one, and each day a covered tool runs without a valid bias audit counts as a separate violation. Missed notices count separately. A non-compliant requisition open for a quarter is not one violation.

Enforcement has been light so far, and that is changing. In December 2025 the New York State Comptroller audited the Department of Consumer and Worker Protection's enforcement and found it ineffective: 75 percent of test calls to the city's 311 hotline about AEDT issues were misrouted and never reached the department, and while the department reviewed 32 posted bias audits and flagged one issue, the Comptroller's review of those same audits identified at least 17 potential violations. The department committed to tighter procedures, so relying on nobody looking is a worse bet than it was.

Which states have AI hiring laws?

As of July 2026, the jurisdictions with AI-specific employment rules in force are New York City, Illinois, California and Texas, with Colorado joining on January 1, 2027. Other states apply general anti-discrimination law to AI-driven decisions without an AI-specific statute. Many more bills are pending.

There is also a live federal fight over whether these state laws survive at all. A December 11, 2025 executive order, "Ensuring a National Policy Framework for Artificial Intelligence," directed the Attorney General to stand up a litigation task force to challenge state AI laws. That effort is itself expected to be challenged, and preemption without an act of Congress is contested. Treat the current state map as authoritative until a court says otherwise.

Can AI make the final hiring decision?

Legally, no rule squarely forbids it in most states, but it is the wrong design. Colorado's new law gives candidates a right to request meaningful human review of an adverse automated decision. Illinois and federal law hold you liable for discriminatory effects regardless of intent. An auto-reject with no human in the loop and no written rationale is the hardest possible fact pattern to defend.

The defensible pattern is narrow and explain, then decide. The system scores candidates against structured, job-related criteria, writes down why each one scored as it did, and hands a ranked shortlist to a person who makes the call. That is how AI candidate screening should work, and it is the principle we cover in more depth in our post on AI recruiting bias and fair screening.

What is a bias audit, and do I need one?

A bias audit is an independent evaluation of a tool's selection rates across demographic categories, expressed as impact ratios. Under NYC's rules it must be conducted by an independent auditor, must be no more than a year old, and a summary must be posted publicly. Outside New York City, no statute currently mandates one, but adverse impact testing is still the best evidence you have that your process is defensible.

Run the numbers anyway. Compare the demographic composition of your applicant pool against your shortlist. The four-fifths rule from the federal Uniform Guidelines on Employee Selection Procedures, which flags any group whose selection rate falls below 80 percent of the highest group's rate, is the standard heuristic and the one the NYC rules build on.

The compliance checklist

• Inventory every tool that touches a hiring decision. ATS knockout rules, resume parsers, ranking features, and any model your vendor quietly shipped last quarter. You cannot audit what you have not listed, and most teams find the list is longer than they expected once they start tracking obligations and controls across the business in one place.
• Map your candidate geography, not your office geography. Local Law 144 follows NYC residents. Illinois follows Illinois positions. Remote roles pull you into jurisdictions you have no office in.
• Write the criteria down before you screen. Job-related, business-necessary, applied identically to every candidate. This is the substance of the defense.
• Disclose AI use to candidates, everywhere. One clear sentence before evaluation, plus the 10 business days of advance notice for NYC candidates.
• Get consent where consent is required. Illinois video interview analysis is the clearest case, and it also requires you to explain how the AI works and what it evaluates.
• Commission the bias audit if NYC covers you. Independent auditor, refreshed annually, summary posted publicly, with selection rates and impact ratios.
• Test for adverse impact even where you are not covered. Compare shortlist composition against applicant pool composition every quarter.
• Demand a written rationale per candidate. A score with no reasoning is a liability with a number attached. If your vendor cannot produce one, that is your answer.
• Keep the records. Four years for California ADS data, three years under Colorado's new law once it lands. Build to the longest requirement that touches you.
• Keep a human deciding, with authority to override, and give candidates a way to ask for that review. A reviewer who rubber-stamps the model's output is not meaningful human review, and Colorado will require a reconsideration route anyway.

Where the law is genuinely unsettled

Be honest with your executives about what nobody knows yet. Whether the December 2025 preemption push succeeds is unresolved. Whether federal deprioritization of disparate impact reduces employer risk is doubtful, since private plaintiffs and state agencies are not bound by an agency's enforcement priorities. Illinois has not finalized its notice rules. And Local Law 144's central term, whether a tool "substantially assists" a decision, has never been tested in a contested case. None of this is a reason to wait: every checklist item above is good practice under any version of the rules that wins.

Screening you can actually defend

The through-line across all of these laws is the same: structured, job-related criteria, applied consistently, disclosed to candidates, explainable after the fact, and decided by a person. That is not a legal burden bolted onto hiring. It is a better hiring process that happens to be defensible. HireAgent screens candidates against your written criteria and returns a ranked shortlist where every candidate carries a rationale tied to those criteria, with AI disclosed, consent respected and an audit trail behind each score. A human makes every hire. See how the scoring works on our AI resume screening page.

General information, not legal advice. Confirm your obligations with employment counsel in the jurisdictions where your candidates live.