These duties attach where the CANDIDATE or EMPLOYEE is located, not where your company is. A company headquartered anywhere in the world that interviews someone sitting in Chicago is inside the Illinois rules for that interview. Remote hiring means most employers of any size are exposed to several of these at once, and a single national process has to satisfy the strictest of them.
Every duty here falls on the employer. A vendor saying their tool is compliant is a commercial assurance, not a legal shield, and in an enforcement action the question is what you did.
Federal anti-discrimination law applies everywhere, with or without an AI statute
Title VII, the ADA and the ADEA govern the outcome, not the method. A tool that screens out disabled applicants because it scores speech patterns creates exposure in every state, including the ones with no AI law at all. The state statutes below add notice, consent and audit duties on top of that floor, they do not replace it.
The duty is yours, not the vendor's
Every obligation here falls on the employer. A vendor's compliance claim is a commercial assurance, not a legal shield, and in an enforcement action the question is what YOU did. Get the vendor's evidence in writing, keep it, and treat it as input to your own decision rather than a substitute for one.
Keep the record now, not when someone asks
What tool, what it was used for, which roles, what notice candidates received, what the audit found and who decided. Reconstructing that after a complaint is expensive and unconvincing. Writing it down as you go costs almost nothing.
No state requires you to train your staff on any of this
Worth knowing before you buy anything sold as mandatory. The duties are notice, consent, audit and non-discrimination. Training is how organisations meet them in practice, and it is not itself a legal requirement anywhere in the United States.
The law and its citation come from our 50-State AI Law Atlas. The mapping from what a tool does to which duty attaches is our reading of that law, kept separate so you can disagree with the reading and still rely on the citation.
Illinois
Triggered when: ai analyses recorded video interviews.
Before AI analyses a video interview you must notify the applicant that AI will be used, explain how it works and what characteristics it evaluates, and obtain consent. The applicant may request deletion within 30 days, and you must also destroy copies held by anyone who received them.
The trap. The explanation requirement is the part vendors cannot satisfy for you. If your provider will not tell you what the model evaluates, you cannot explain it, and you therefore cannot lawfully use it on an Illinois applicant.
Illinois
Triggered when: ai screens, scores or ranks applicants, or ai is used in promotion, discipline or performance decisions, or ai substantially assists any employment decision.
Using AI that has a discriminatory effect in recruitment, hiring, promotion, discipline or other employment decisions is prohibited, including through proxies such as ZIP code. Employees must be notified that AI is in use.
The trap. Discriminatory EFFECT, not intent. A tool nobody designed to discriminate still breaches this if the outcomes land unevenly, and the ZIP code proxy is named explicitly because it is the commonest way a model rediscovers race from geography.
Maryland
Triggered when: ai performs facial recognition or facial analysis, or ai analyses recorded video interviews.
You may not use facial recognition or AI image analysis on an applicant's video interview without the applicant's written consent.
The trap. Written consent, specifically, and consent to the interview itself is not consent to the analysis. A general clause buried in an application form is a weak foundation for this one.
New York (incl. NYC)
Triggered when: ai screens, scores or ranks applicants, or ai is used in promotion, discipline or performance decisions, or ai substantially assists any employment decision.
For hiring or promotion in New York City, an automated employment decision tool requires an independent bias audit within the previous year, publication of the audit results, and notice to candidates at least ten business days before use.
The trap. The audit must be independent and no more than a year old, and the results must be published where a candidate can find them. A vendor's own fairness report is not an independent audit, and an audit performed once is out of date twelve months later.
Connecticut
Triggered when: ai screens, scores or ranks applicants, or ai is used in promotion, discipline or performance decisions, or ai substantially assists any employment decision.
Connecticut's omnibus AI act creates staged duties for developers and deployers, including workplace AI, running from 1 October 2026 through 1 January 2028. Confirm which stage applies before your next hiring cycle.
The trap. The staging is the trap. A duty that does not apply this quarter may apply next, and a programme built once against the earliest date will drift out of compliance without anyone touching it.
Connecticut Connecticut Artificial Intelligence Responsibility and Transparency (CART) Act
Omnibus AI act creating an AI Policy Office and learning laboratory, with staged duties (October 1, 2026 through January 1, 2028) covering AI developers and deployers, workplace AI, AI companions, and youth online safety.
SB 5 (P.A. 26-15). Effective 2026-10-01. Checked 2026-08-02.
Illinois Artificial Intelligence Video Interview Act
First-in-the-nation requirement that employers notify, explain to, and obtain consent from job applicants before AI analyzes their video interviews, with a 30-day deletion right.
820 ILCS 42 (P.A. 101-0260). Effective 2020-01-01. Checked 2026-08-02.
Illinois Human Rights Act, Artificial Intelligence in Employment
Amends the Illinois Human Rights Act to prohibit employers from using AI that has a discriminatory effect in recruitment, hiring, promotion, discipline and other employment decisions (including ZIP-code proxies), and requires employee notice of AI use.
HB 3773 (P.A. 103-0804). Effective 2026-01-01. Checked 2026-08-02.
Illinois Teacher Evaluation, AI Restrictions
Restricts use of artificial intelligence in public-school teacher evaluations and requires disclosure and human judgment in evaluation decisions; effective January 1, 2027.
SB 2909 (P.A. 104-0565). Effective 2027-01-01. Checked 2026-08-02.
Maryland Facial Recognition in Job Interviews, Consent
Prohibits employers from using facial-recognition (AI image-analysis) services on applicant video interviews without the applicant's written consent.
HB 1202 (Ch. 533, 2020). Effective 2020-10-01. Checked 2026-08-02.
New York (incl. NYC) NYC Local Law 144, Automated Employment Decision Tools
Requires annual independent bias audits of automated employment decision tools used for hiring or promotion in New York City, plus candidate notice and publication of audit results; DCWP enforcement began July 5, 2023.
Local Law 144 of 2021 (NYC Admin. Code §20-870 et seq.). Effective 2023-07-05. Checked 2026-08-02.
Is it legal to use AI in hiring?
Yes, with conditions that depend on where your candidates are and on what the AI does. Illinois requires notice, explanation and consent before AI analyses a video interview, and separately prohibits AI with a discriminatory effect in employment decisions. Maryland requires written consent for facial recognition in interviews. New York City requires an independent bias audit within the previous year, published results and ten business days notice. Connecticut's duties stage in from October 2026. Federal anti-discrimination law applies everywhere regardless.
Do AI hiring laws apply based on where the company is or where the candidate is?
Where the candidate is. This is the most consequential misunderstanding in the area. A company with no presence in Illinois that interviews a candidate sitting in Chicago is inside the Illinois rules for that interview. Because remote hiring is normal, most employers of any size are exposed to several of these regimes at once, and one national process has to satisfy the strictest.
What is a bias audit for an AI hiring tool?
Under New York City's Local Law 144, an automated employment decision tool used for hiring or promotion in the city needs an independent bias audit conducted within the previous year, with the results published and candidates notified at least ten business days before use. Independent is the operative word: a vendor's own fairness report does not satisfy it, and an audit is out of date a year after it was performed.
Our vendor says their tool is compliant. Is that enough?
No. Every one of these duties falls on the employer, so a vendor's assurance is a commercial promise rather than a legal shield. In an enforcement action the question is what you did: what notice you gave, what consent you obtained, what audit you held and what you knew about how the tool scores people. Get the vendor's evidence in writing and treat it as input to your own decision.
Does AI in hiring have to be disclosed to candidates?
In several places, yes, and the form differs. Illinois requires notification and an explanation of what the AI evaluates before a video interview, plus consent. Illinois also requires employee notice where AI is used in employment decisions. New York City requires candidate notice at least ten business days before using an automated employment decision tool. Maryland's requirement is written consent specifically for facial recognition.
Information, not legal advice. Obligations turn on facts this page cannot see, including how your tool actually scores people, and the law is moving. Every citation is linked so you can check it, which is the only way any of this should be used.