AI in Hiring Laws: What Small Employers Need to Know in 2026

Your applicant tracking system ranks resumes and a chatbot asks the screening questions. When AI decides who moves forward, AI in hiring laws apply. As of October 2026 there’s no single federal AI statute, but federal anti-discrimination law covers these tools, and New York City, Illinois, California, Texas and Maryland add rules. Colorado’s rules start on January 1, 2027.

Which AI in hiring laws apply, by state?

Short answer: Federal anti-discrimination law applies to AI hiring tools everywhere. On top of it, New York City requires bias audits and notices, Illinois requires notices and bans discriminatory AI use, California’s regulations cover automated systems, Texas bars intentional AI discrimination, Maryland requires consent for facial recognition in interviews, and Colorado’s rules start in 2027.

If you’d rather have people, not software, run the first screen, our recruitment process outsourcing team screens candidates and you make the hire.

Federal guidance has shifted too. The Equal Employment Opportunity Commission (EEOC) has taken its AI guidance off eeoc.gov, including its ADA document and its May 2023 Title VII document on adverse impact. As of October 2026, the old addresses return “Page not found.” And two executive orders, covered below, changed enforcement priorities but not the statutes.

What AI hiring tools do small employers actually use?

Short answer: Small employers, as we see it, mostly use AI features built into tools they already pay for: resume ranking in an applicant tracking system (ATS), screening chatbots, interview scheduling and, less often, one-way video interviews or online assessments with automated scoring.

The Census Bureau’s Business Trends and Outlook Survey found that between 17% and 20% of businesses used AI in some business function from December 2025 to May 2026, and fewer than 20% of firms with four or fewer employees did. For a small employer, AI usually arrives as a switch someone turned on in the hiring software, not as a deliberate purchase.

A tool that only helps with admin work carries little risk. One that scores, ranks, filters or rejects people is where every law in this guide applies. If a chatbot drafts your job ads, the wording still has to be lawful; our guide on how to write a job description has a wording check.

What does NYC Local Law 144 require?

Short answer: New York City’s Local Law 144 covers employers that use an automated employment decision tool (AEDT) for a city-linked job. You need an independent bias audit from the past year and a public summary of the results. You also need to notify candidates who live in the city at least 10 business days before you use the tool.

A tool counts when it uses machine learning, statistical modeling, data analytics or AI, helps make employment decisions, and “substantially assists or replaces” human judgment, according to the Department of Consumer and Worker Protection’s FAQ. The law applies when the job is at a New York City office at least part time, when a fully remote job is associated with a city office, or when the employment agency using the tool is in the city.

What compliance looks like:

• The audit: an impartial evaluation by an independent auditor that, at a minimum, calculates selection or scoring rates and impact ratios by sex, by race and ethnicity, and for intersectional categories. A vendor can commission the audit of its own tool.
• The summary: posted on the employment section of your website, or linked from it.
• The notice: in the job posting, by mail or email, or on your careers page. It must name the job qualifications or characteristics the tool will assess and explain how to ask for a reasonable accommodation.

Watch out: A vendor’s audit doesn’t move the duty off you. The city’s FAQ says employers and employment agencies are “ultimately responsible” for making sure a bias audit was done before they use the tool.

Enforcement has been light so far, and that may change. A December 2, 2025 audit by the State Comptroller found the city’s complaint process ineffective. The agency had reviewed 32 companies and found one issue; the Comptroller’s team reviewed the same companies and found at least 17 instances of potential non-compliance. The same report notes the city can impose civil penalties between $500 and $1,500 per day.

What do Illinois’ AI hiring laws require?

Short answer: Illinois has two AI hiring laws. The Artificial Intelligence Video Interview Act covers AI analysis of recorded interviews for Illinois-based jobs, and since January 1, 2026 the Illinois Human Rights Act bars AI use that discriminates and requires employers to tell employees when they use AI in employment decisions.

The AI Video Interview Act has applied since January 1, 2020. Before an applicant records a video interview that AI will analyze, you must tell them AI may be used, explain how it works and what general characteristics it evaluates, and get their consent. You can’t use AI on applicants who don’t consent. Share the videos only with people needed to evaluate the applicant. If asked, delete them, backups included, within 30 days. If you rely only on AI to pick who gets an in-person interview, you also report applicants’ race and ethnicity to the state each year.

The newer rule comes from Public Act 103-0804 (HB 3773). It makes three things civil rights violations:

• using AI “that has the effect of” discriminating in recruitment, hiring, promotion, discipline, discharge or other terms of employment;
• using ZIP codes as a proxy for protected classes;
• failing to give notice that you use AI for those purposes.

The test is the effect, so you can break this rule without meaning to. According to the Illinois Human Rights Commission, it covers employers with one or more Illinois employees for 20 or more weeks a year, and applicants and employees can file administrative charges and then sue.

Proposed rules, described in the Commission’s April 2026 presentation, would require the notice to name the AI product, its developer and vendor, the decision it influences, its purpose and data categories, the job types, a contact person and how to request an accommodation. As of October 2026 we couldn’t confirm they’re final, so treat that list as the safe minimum.

Is the Colorado AI Act in effect?

Short answer: No, the Colorado AI Act isn’t in effect. The 2024 law (Senate Bill 24-205) never took effect. On May 14, 2026, Colorado’s governor signed Senate Bill 26-189, which repealed and re-enacted it as narrower rules for automated decision-making technology that take effect January 1, 2027.

The new law applies to decisions made on or after January 1, 2027, and employment is one of its covered areas. If you use a covered tool to “materially influence” a hiring decision, you’ll owe three things:

1. Notice before use. A clear notice at the point of interaction, or a prominent public notice near where the decision happens, saying you use the tool and how to get more information.
2. An explanation after a rejection. Within 30 days of an adverse outcome, a plain-language description of the decision and the tool’s role, how to request details such as the tool’s name, version, developer and data used, and the person’s rights.
3. Correction and human review. People can ask to correct factually wrong personal data and request meaningful human review and reconsideration.

The law leaves out low-stakes tasks such as routine scheduling and workflow management, and tools that only summarize or organize information for a person without producing a score, ranking or recommendation. The Attorney General enforces it and must adopt rules on the adverse-outcome disclosures by January 1, 2027. Until January 1, 2030, the Attorney General must give a 60-day notice and a chance to cure before bringing an action. It doesn’t create a new private right to sue.

What do Title VII and the ADA mean for AI hiring tools in 2026?

Short answer: Title VII of the Civil Rights Act of 1964 and the Americans with Disabilities Act (ADA) apply to AI hiring tools in full. If you have 15 or more employees, an AI screen that disproportionately filters out a protected group, or screens out people with disabilities, is unlawful unless you can show it’s job related and consistent with business necessity.

Title VII covers practices that cause a disparate impact by race, color, religion, sex or national origin, with that same job-related, business-necessity defense. The ADA bars tests and selection criteria that screen out people with disabilities on the same terms, and requires tests to be given in a way that measures skills rather than impairments. Neither law cares whether the test is a paper form or a model.

Federal enforcement has shifted, though. Executive Order 14281, signed April 23, 2025, tells federal agencies to deprioritize disparate-impact enforcement and names Title VII’s disparate-impact section. It also directed the EEOC to review pending matters built on that theory. The statute’s text didn’t change, and state laws such as Illinois’ reach discriminatory effects on their own terms.

Executive Order 14365, signed December 11, 2025, set up a Justice Department task force to challenge state AI laws and asked for a federal framework that would preempt them. It directs federal agencies and doesn’t cancel any state law by itself, so keep following the state rules that apply to you.

Private lawsuits haven’t waited for the agencies. In Mobley v. Workday, in federal court in Northern California, an applicant alleges the vendor’s AI recommendation system discriminated by race, age and disability. On May 16, 2025 the judge granted preliminary certification of a collective action on the age claim for applicants 40 and over. As of October 2026 the case is pending, with a class certification motion set for hearing on March 9, 2027.

Our take: Don’t read the federal pullback as permission. Applicants can still bring claims, and states keep writing their own AI rules. Build your process to survive a disparate-impact question.

Which other states have AI hiring rules?

Short answer: Beyond New York City, Illinois and Colorado, three more states have AI hiring rules in force that we could confirm on official pages: California, Texas and Maryland. California’s regulations, in force since October 1, 2025, are broad. Texas’ House Bill 149, in force since January 1, 2026, is narrow and requires intent. Maryland’s 2020 law covers one tool: facial recognition in job interviews, which needs the applicant’s signed waiver.

California. The Civil Rights Council’s regulations took effect October 1, 2025, according to the Civil Rights Department. They define an automated-decision system broadly: resume screening for terms or patterns, analysis of facial expression, word choice or voice in online interviews, puzzles and games that measure aptitude or “cultural fit,” and directing job ads to targeted groups. In a discrimination claim, evidence of anti-bias testing, or the lack of it, is relevant, including how good, recent and broad it was. Employers must keep these records for four years, and assessments that draw out disability-related information may be an unlawful medical inquiry.

California’s privacy agency adds separate automated decisionmaking rules: from January 1, 2027, businesses that use the technology for significant decisions such as employment must give notice, offer an opt-out where it applies and explain how the tool worked. These apply only to businesses covered by the state privacy law, such as those with annual gross revenue above $26,625,000 (the 2025 figure).

Texas. HB 149, signed June 22, 2025 and in force since January 1, 2026, bars developing or deploying an AI system with the intent to unlawfully discriminate against a protected class. It states that a disparate impact alone isn’t enough to show intent.

Maryland. Since October 1, 2020, Md. Code, Lab. & Empl. § 3-717 has barred employers from using a facial recognition service to create a facial template during an applicant’s interview unless the applicant consents by signing a waiver (it came from HB 1202 of 2020). The waiver must state, in plain language, the applicant’s name, the interview date, that they consent to facial recognition during the interview and whether they read the waiver. If a video-interview tool analyzes faces, get the waiver signed before the interview.

A checklist for using AI recruiting tools lawfully

Work through this list once for every tool, then repeat it each year or whenever a vendor ships a major update. It’s a summary, not legal advice; where a rule could cost you real money, have an employment lawyer look at your setup.

1. Inventory the tools. List everything that touches applicants and ask each vendor which features use AI and whether any of them score, rank or reject people.
2. Map your jobs and candidates. Note any role tied to New York City, Illinois, California, Maryland (if a video tool reads faces) or, from 2027, Colorado. Remote roles count where the candidate or the associated office is.
3. Get the vendor’s paperwork. Ask for the latest bias audit and its impact ratios, what data the tool uses and how it’s validated for the job. Put audit cooperation and video deletion in the contract.
4. Run your own selection-rate check. Under the federal four-fifths rule, a selection rate for any group below 80% of the highest group’s rate is generally treated as evidence of adverse impact.
5. Send the notices. New York City: 10 business days before use, with accommodation instructions. Illinois: notice of AI use, plus consent for video analysis. Maryland: a signed waiver before facial recognition in an interview. Colorado and California’s privacy rules from 2027.
6. Offer another path. Tell candidates how to ask for an accommodation or a human review, and make sure someone answers.
7. Keep a person in the decision. Nobody should be rejected by software alone. The reviewer needs the authority to overrule the tool, and the time to do it.
8. Remove proxies. Turn off filters on ZIP code, graduation year, employment gaps and similar fields unless you can show they’re needed for the job.
9. Keep records for four years. That meets California’s rule, the longest among the laws here.

Example: Say a screening chatbot advances 50 of 100 male applicants (50%) and 30 of 100 female applicants (30%). Thirty divided by 50 is 60%, well under the four-fifths line, so you’d pause the tool, find the question causing the gap and fix it before using it again.

Where to start this week

1. Turn off auto-reject in every tool you use until you’ve reviewed it. Let software sort, and let a person decide.
2. Ask each vendor three questions: what the AI does, where the latest bias audit is, and what data it uses.
3. Add an AI notice and accommodation line to your careers page and job template if you hire in New York City or Illinois.
4. Put January 1, 2027 on the calendar for Colorado and California’s privacy rules, and revisit your notices before then.
5. Pick one person who owns the checklist and reviews it every year.

For the pay rules that sit alongside these, see our guide to pay transparency laws by state. Our diversity recruiting service covers widening the pool without shortcuts.