AI Hiring Laws by State: Is AI Resume Screening Legal in 2026?

Short answer: yes, AI resume screening is legal in the United States. No federal law bans it, and no state bans it. What the law regulates is how you use it. Federal anti-discrimination statutes (Title VII, the ADA, the ADEA) apply to an algorithm exactly as they apply to a human reviewer, and a growing set of state and city laws add specific duties on top: bias audits and candidate notice in New York City, notice and an explicit discrimination ban in Illinois, recordkeeping and vendor diligence in California, and governance obligations in Colorado from 2027. The employer, not the vendor, carries the liability.

Last updated July 2026. This is a plain-language summary for hiring teams, not legal advice. AI employment law is moving quickly and several dates below have already been changed once by legislators, so confirm the current position with counsel before you rely on it.

Is AI resume screening legal?

It is legal, and it is widely used. Screening software that reads resumes, applies knockout criteria, scores role-fit and ranks a pile of applicants is lawful in every US state today. There is no statute that prohibits using AI to help decide who gets an interview.

The catch is that "legal" does not mean "unregulated." Every anti-discrimination rule that governs a human recruiter governs the software too, and the employer using the tool is the one who answers for the outcome. If a screening tool filters out candidates over 40, or people with disabilities, or one racial group at a much higher rate than another, the fact that a model produced the result is not a defense. Courts and agencies treat the tool as the employer's chosen selection procedure. That principle has not changed, and it is the single most important thing to understand before you buy anything.

What changed at the federal level in 2025 and 2026

Here is where a lot of published advice is now out of date. In May 2023 the EEOC issued technical assistance explaining how Title VII and the ADA apply to AI-driven selection. On January 27, 2025, the EEOC removed that AI guidance from its website, following Executive Order 14179 and a broader federal shift away from AI regulation. Articles that tell you to "follow EEOC AI guidance" are pointing at documents the agency has taken down.

Read that carefully, because the conclusion people jump to is wrong. The guidance was withdrawn. The law was not. Title VII, the ADA and the ADEA are statutes passed by Congress, and no executive order repeals them. Disparate impact liability still exists. The four-fifths rule (a selection rate for one group below 80 percent of the top group's rate) is still the standard first test that plaintiffs' lawyers, auditors and courts reach for. Private lawsuits and state enforcement did not pause. What actually happened is that the federal government stopped explaining how it would apply the law, while the states started writing rules of their own. For an employer, that is a harder environment to operate in, not an easier one.

AI hiring laws by state, July 2026

Two patterns are worth pulling out of that table. First, the trend line is toward notice: most of these laws, whatever else they do, require you to tell candidates that AI is involved. Second, only New York City actually forces you to commission and publish an independent bias audit. That audit is the expensive obligation, and it is the one buyers most often discover after they have signed.

Does NYC Local Law 144 apply to my company?

It applies if you use an automated employment decision tool to substantially assist a hiring or promotion decision for a job located in New York City, or for a candidate who is an NYC resident, regardless of where your company is headquartered. A Chicago employer hiring for a Manhattan office is covered. Penalties run from $500 for a first violation to $1,500 for subsequent ones, and each day of non-compliant use can count separately, so the exposure compounds rather than capping out.

The enforcement climate has also shifted. In December 2025 the New York State Comptroller published an audit of how the Department of Consumer and Worker Protection had enforced the law between July 2023 and June 2025, and the agency committed to a more rigorous approach going forward. The early era of "the law exists but nobody is checking" is closing. Our guide to the Local Law 144 bias audit covers what the audit itself involves.

Can AI reject candidates automatically?

Legally, in most of the country, nothing stops it. Practically, it is the design decision most likely to get you sued, and we think it is the wrong one regardless of the law. Automated rejection turns a screening tool into the decision maker, which is exactly the posture regulators, plaintiffs and juries respond to worst. It also removes the human who would have caught the qualified candidate the model misread, the career-changer whose experience does not use your keywords, and the applicant whose disability accommodation request never reached a person.

Keeping a recruiter in the loop is the cheapest risk control available. A tool that ranks and flags, and leaves every rejection to a human, keeps the employer in the position the law has always assumed: a person made the call, and can explain why. Prescreener is built this way on purpose. It ranks the pile and shows the reasons, and it never auto-rejects anyone.

What is a bias audit and who can run one?

A bias audit is an independent statistical review of a screening tool's outcomes. It calculates selection rates by race and ethnicity, and by sex, then expresses them as impact ratios against the most-selected group, using the four-fifths rule as the reference threshold. Under Local Law 144 it must be performed by an auditor who is genuinely independent of the vendor and the employer, repeated annually, and summarized publicly on your careers site.

An audit tells you whether outcomes look skewed. It does not tell you why, and it does not make a skewed tool lawful. If your impact ratio is poor, the follow-up question is whether the criterion driving it is job-related and consistent with business necessity, which is a legal and operational question rather than a statistical one.

How to use AI screening without breaking the law

Six things do most of the work, and none of them require a legal department:

• Screen on job-related criteria only. Every knockout should be something you could defend out loud as necessary for the job. Certifications, licenses, shift availability, work authorization: fine. Proxies for age, race, national origin or disability, including graduation year and zip code: not fine, and Illinois now names zip code explicitly.
• Apply the same criteria to everyone. Consistency is both the fairness argument and the efficiency argument. It is also what an auditor will test.
• Disclose that AI is used. Notice is the common thread across NYC, Illinois and Colorado, and it costs you nothing. Put it in the job posting and the application flow.
• Keep a human in every rejection. Rank and flag, never auto-reject.
• Keep the records. California expects four years. Retention is cheap; reconstructing a screening decision from memory two years later is not.
• Ask the vendor for the audit, in writing. If a vendor cannot show you bias audit results and cannot explain what its tool used to rank a candidate, the compliance risk lands on you, not on them.

Public employers carry an extra layer, because their hiring decisions are also reviewable by a board, a union, or an open records request. School districts are the clearest example: the criteria applied to each posting have to be written down and defensible before the spring wave arrives, which is a large part of why the teacher hiring process is built around a documented screening standard per classification.

Treat that list as a control set you own rather than a form you file once. Hiring teams that already track their obligations and controls in one place tend to have a much easier time when a regulator, an auditor or an opposing counsel asks how a specific candidate was screened, because the answer is written down rather than reconstructed.

What this means when you are choosing a screening tool

The buying question is not "is this legal," because the honest answer for nearly every product on the market is yes. The question is whether the vendor's design makes your compliance easier or harder. Three things tell you quickly: whether it explains its rankings, whether it auto-rejects, and whether it will hand you bias audit results without a fight. Once you have picked a tool, our practical checklist on reducing bias in AI resume screening covers how to configure and audit it so it stays defensible.

Prescreener reads every inbound resume, applies the knockout and eligibility criteria you set, asks screening questions, scores role-fit, and ranks the pile with a transparent reason attached to each candidate. Screening is bias-audited to support EEOC standards and NYC Local Law 144, candidates are told AI is used, and a recruiter makes every decision. If you want the mechanics of how that works in practice, the AI applicant screening page walks through it, the candidate screening software page covers the ranking and criteria setup, and our explainer on EEOC standards and AI goes deeper on adverse impact and accommodation.