New Jersey AI hiring guidance (2026): the Division on Civil Rights guidance, N.J.A.C. 13:16 and pending bills
On this page
- What New Jersey rules apply to AI in hiring, and what is their status?
- What does the Division on Civil Rights guidance say?
- What do the N.J.A.C. 13:16 rules add for hiring tools?
- Who can bring a claim, and how fast?
- Which New Jersey AI hiring bills are pending?
- Does the New Jersey Data Privacy Act cover job applicants?
- What should recruiters and staffing agencies in New Jersey do now?
- Questions people ask
New Jersey has no AI-specific hiring statute as of November 2026. It regulates AI hiring tools through the Law Against Discrimination (LAD, N.J.S.A. 10:5-1 et seq.): the Division on Civil Rights (DCR) issued guidance on algorithmic discrimination on January 9, 2025, and its disparate impact rules at N.J.A.C. 13:16, effective December 15, 2025, define "automated employment decision tools" and say an untested tool may have an unlawful disparate impact. Employers are liable for a biased tool even if a vendor built it and even without intent to discriminate. Bills that would add audits, notices and impact assessments, led by S4075, are pending and are not law.
The state's other hiring rules are summarized on hiring employees in New Jersey, and what changes in 2027 is on new hiring laws in New Jersey for 2027. This page covers only AI and automated tools: what the guidance says, what the rules add, who can bring a claim, and the status of each bill.
Not legal advice. This summarizes the DCR's January 2025 guidance, N.J.A.C. 13:16 as adopted by R.2025 d.150, the DCR's March 2026 disparate impact FAQs, and bill records on the New Jersey Legislature's site, as checked on November 17, 2026. Ask New Jersey employment counsel how they apply to your tools.
What New Jersey rules apply to AI in hiring, and what is their status?
| Rule | Who it covers | What it requires | In force from | Source |
|---|---|---|---|---|
| Law Against Discrimination | Employers of any size, employment agencies, labor organizations | No disparate treatment, disparate impact or failure to accommodate, whatever tool is used | In force (enacted 1945) | N.J.S.A. 10:5-5, 10:5-12 |
| Guidance on Algorithmic Discrimination | All LAD covered entities | Explains how the LAD applies to automated decision-making tools; imposes no new duties | Issued January 9, 2025 | NJ Attorney General |
| Disparate impact rules, N.J.A.C. 13:16 | Employers, labor organizations, employment agencies and other covered entities | Burden-shifting test, empirical evidence, vendor reasonable steps, automated tool examples, Uniform Guidelines applied to all LAD traits | December 15, 2025 | Adoption announcement |
| S4075 and A4981 (AI monitoring and decision systems) | Employers using such systems | Impact assessment approved by an independent auditor, notices, human oversight | Pending, not law | S4075 |
| A2726, A1021, S3263 | Employers and tool vendors | Tool regulation, bias audit standards, AI video interviews | Pending in committee, not law | NJ Legislature bill pages (below) |
What does the Division on Civil Rights guidance say?
The guidance defines an "automated decision-making tool" as any technological tool, "including but not limited to, a software tool, system, or process that is used to automate all or part of the human decision-making process," from generative AI to decision trees. Its core point is that the LAD "draws no distinctions based on the mechanism of discrimination." A covered entity "can violate the LAD even if it has no intent to discriminate, and even if a third-party was responsible for developing" the tool. It names three ways a tool can break the law:
- Disparate treatment. Designing or using a tool to treat a protected class differently, using it selectively on one group, or relying on a close proxy for a protected trait.
- Disparate impact. A tool whose recommendations disproportionately harm a protected class is unlawful unless it serves a substantial, legitimate, nondiscriminatory interest and no less discriminatory alternative exists. Whether the employer tested the tool for bias "may be considered as relevant evidence."
- Reasonable accommodation. A tool that blocks accommodation for disability, religion, pregnancy or breastfeeding. The hiring example is a typing-speed test that cannot measure typing from a non-traditional keyboard; another is a tool trained without data from people who use accommodations, which may screen out applicants who could do the job with one.
The guidance also says an entity is not shielded "because the tool was developed by a third party or because the entity does not understand the inner workings of the tool." Its footnote 51 lists steps covered entities can take: quality control of training data, impact assessments, pre- and post-deployment bias audits by independent parties, notice that a tool is used, involving affected people in development, and red-teaming. The DCR presents these as options, not requirements, and the guidance states that it "does not impose any new or additional requirements."
What do the N.J.A.C. 13:16 rules add for hiring tools?
The rules, adopted by R.2025 d.150 and effective December 15, 2025, turn the disparate impact standard into regulation text. Section 13:16-1.3 defines automated employment decision tools as "any software, system, or process that aims to automate, aid, or replace human decision-making relevant to employment," including tools that generate "scores, rankings, predictions, classifications, or some recommended action(s)" used for "advertising, recruiting, screening, interviewing, hiring, placement, promotion, and compensation." Because the definition includes tools that only "aid" a decision, it reaches tools whose output a person reviews before deciding, not just tools that decide on their own.
Section 13:16-3.2(c) gives three hiring examples:
- Tools trained on your current staff. A tool that learns from a mostly white, cisgender male workforce may score women lower for listing "women's field hockey" or Black applicants lower for listing a "Black Student Alliance."
- Schedule screens. A tool that screens out applicants based on their schedule may have a disparate impact based on religion, disability or medical condition "and must include a mechanism for applicants to request a reasonable accommodation." This is the one place the rules use "must" about a tool's design.
- Untested tools. A tool "that has not been adequately tested and shown to not adversely affect people in a protected class before its use may have a disparate impact." The example is facial analysis that scores personality in video interviews and was tested mainly on white people without disabilities, or not on people who wear religious headwear or facial hair.
Four other sections matter for anyone using a vendor's tool:
- Burden of proof (13:16-2.2). The applicant shows a disparate impact; the employer then shows the practice is "job related and consistent with a legitimate business necessity"; the applicant can still win by showing a less discriminatory alternative. Every step needs "empirical evidence," such as applicant selection rates.
- Vendors (13:16-2.4(e)). If a practice relies on "products, procedures, or systems of an outside person or vendor," the employer "must take reasonable steps to ensure" they are consistent with the LAD.
- Uniform Guidelines for every LAD trait (13:16-3.3). The federal Uniform Guidelines on Employee Selection Procedures (29 CFR part 1607) are incorporated and applied to all LAD protected characteristics, which include age, sexual orientation, gender identity and genetic information as well as the Title VII traits. Our four-fifths rule guide explains the selection-rate check they use.
- Agencies and timing (13:16-3.1(c), 2.1(b)). The employment rules apply to employment agencies. A practice can be challenged once it is "approved, announced, or otherwise finalized," before it is used, but not while it is still under internal debate.
The DCR's March 2026 FAQs repeat that the rules "do not create new liability or change the law"; they codify standards the courts already apply. The federal side of the same analysis is on Title VII disparate impact and AI hiring.
Who can bring a claim, and how fast?
The LAD's definition of "employer" includes "all persons" unless specifically exempt, so there is no minimum employee count. Under N.J.A.C. 13:16-2.1(c), an aggrieved person, the Attorney General, the DCR Director and certain other officials may file a disparate impact complaint with the DCR or sue in Superior Court. The DCR's complaint page says its intake checks whether the violation "occurred within the past 180 days," and a complainant may move to Superior Court "so long as you are within the 2-year statute of limitations." The DCR can order compensatory damages, damages for pain and humiliation, attorneys' fees, reinstatement and new policies.
Which New Jersey AI hiring bills are pending?
None of these has passed either house as of November 17, 2026, according to the Legislature's bill histories:
| Bill | What it would do | Status (2026-2027 session) |
|---|---|---|
| S4075 | Regulates AI electronic monitoring tools and automated employment decision systems: an impact assessment approved by an independent auditor before use, 60 days' notice to employees and unions, no decisions based exclusively on system outputs, designated human reviewers, appeal rights; takes effect 18 months after enactment | Introduced May 4, 2026; reported by the Senate Labor Committee October 1, 2026; referred to Senate Budget and Appropriations; fiscal estimate dated November 5, 2026 |
| A4981 | Assembly bill with the same synopsis as S4075 | Introduced May 7, 2026; in the Assembly Science, Innovation and Technology Committee |
| A2726 | "Regulates use of automated employment decision tools in employment decisions to minimize discrimination in employment" | Introduced January 13, 2026; in committee |
| A1021 | "Creates standards for independent bias auditing of automated employment decision tools" | Introduced January 13, 2026; in committee |
| S3263 | "Regulates use of artificial intelligence-enabled video interview in hiring process" | Introduced February 2, 2026; in the Senate Labor Committee |
The S4075 summary above comes from the Senate Labor Committee's October 1, 2026 statement. The bill covers applicants: its definition of an automated employment decision system reaches outputs "used by an employer to make employment-related decisions about an employee or applicant for employment." Do not build policies around any of these bills yet; check the bill pages before you renew an AI screening contract.
Does the New Jersey Data Privacy Act cover job applicants?
The Act (P.L.2023, c.266) lets consumers opt out of profiling for decisions with legal or similarly significant effects, and those decisions include "employment opportunities." But the statute says a consumer "shall not include a person acting in a commercial or employment context." On its face that leaves job applicants dealing with an employer outside the definition, so unlike California's CCPA automated decisionmaking rules, New Jersey's privacy law is not the main constraint on AI screening. The LAD is.
What should recruiters and staffing agencies in New Jersey do now?
- List every tool that scores, ranks, filters or recommends at any stage, including ad targeting, knockout questions, assessments and interview scoring. Under the 13:16 definition, tools that only "aid" a person count.
- Test before use and keep testing. Track selection rates by protected class at each automated step, using the Uniform Guidelines method, and keep the results. The rule's example treats an untested tool as a likely disparate impact problem.
- Put an accommodation path in every automated step. A schedule screen must have one, and the guidance's typing-test example shows why other assessments need one too. See accommodations in interviews.
- Remove proxies. Ask what inputs drive scores; names, zip codes, graduation years, club names and gaps can stand in for protected traits.
- Write vendor duties into the contract: testing data by protected class, notice of model changes, help answering a DCR complaint, and a way to switch features off. That is how you show the "reasonable steps" N.J.A.C. 13:16-2.4(e) requires.
- Staffing agencies: the rules apply to you as an employment agency. Do not run a client's automated screen you could not defend yourself, and agree in writing who tests it.
- Watch the bills. If S4075 or A2726 is enacted, notice and audit duties would follow; New Jersey candidates for New York City roles are already covered by Local Law 144.
For how New Jersey compares with California, Colorado, Illinois, Texas and other states, see AI hiring laws by state.
Questions people ask
Does New Jersey have a law regulating AI in hiring?
Not a separate AI statute, as of November 2026. New Jersey regulates AI hiring tools through its Law Against Discrimination: the Division on Civil Rights issued guidance on algorithmic discrimination on January 9, 2025, and its disparate impact rules at N.J.A.C. 13:16, effective December 15, 2025, define automated employment decision tools and give hiring examples. Bills that would add audits, notices and impact assessments, such as S4075, are pending and not law.
Can a New Jersey employer be liable for a biased AI tool it bought from a vendor?
Yes. The Division on Civil Rights guidance says a covered entity is not shielded from liability because a third party developed the tool or because the entity does not understand how it works, and N.J.A.C. 13:16-2.4(e) requires an employer whose practice relies on a vendor's products or systems to take reasonable steps to ensure they are consistent with the Law Against Discrimination.
Do New Jersey employers have to audit AI hiring tools for bias?
No rule in force requires a formal bias audit, as of November 2026. But N.J.A.C. 13:16-3.2(c)(3) says using an automated employment decision tool that has not been adequately tested and shown not to adversely affect a protected class may have a disparate impact, and the guidance says testing for bias is relevant evidence. Pending bills A1021 and S4075 would add audit or impact assessment duties if enacted.
Does the New Jersey Data Privacy Act give job applicants an AI opt-out?
The Act's profiling opt-out is written for consumers, and the statute says a consumer does not include a person acting in a commercial or employment context. Job applicants dealing with an employer therefore fall outside its definition on the statute's face, and the Law Against Discrimination remains the main rule for AI in New Jersey hiring.
What is New Jersey bill S4075?
S4075 would regulate AI-based electronic monitoring tools and automated employment decision systems, requiring an impact assessment approved by an independent auditor before use, 60 days' written notice to employees, human oversight and appeal rights. The Senate Labor Committee reported it on October 1, 2026 and referred it to the Senate Budget and Appropriations Committee; as of November 17, 2026 it had not passed either house.