An AI Rejected Your Résumé. New Jersey Won’t Say Why.

Derek Mobley’s case began with a familiar object: the automated rejection notice. Mobley, a Black applicant over 40 who has identified anxiety and depression as disabilities, alleged in federal court that employers using Workday rejected him repeatedly, sometimes soon after he submitted an application. His claims remain allegations, and the litigation has not established that Workday’s software discriminated against him.

The case matters because Mobley did not receive the thing applicants tend to want first. He did not get a legible account of what the system read, how it scored him or which qualification moved his application into the discard pile. He had a pattern of rejections and a vendor’s name. To obtain more, he had to convert that pattern into a discrimination lawsuit.

A federal judge ruled in 2024 that Mobley had adequately alleged, at the pleading stage, that Workday could act as an employer’s agent under federal anti-discrimination law. That was an important opening. It was not a right to an explanation.

For an applicant in New Jersey, the same automated notice sits at the center of a jurisdictional problem. The relevant law depends on where the job is based, what the software did, what data fed it and whether the applicant can identify discrimination or an inaccurate consumer report. No single regulator owns the rejection.

New

Jersey gives you a discrimination claim, not a receipt

The New Jersey Law Against Discrimination applies when an employer uses software as well as when a manager makes the decision personally. An employer cannot insulate a discriminatory hiring practice by buying it from a vendor. New Jersey’s Division on Civil Rights has also made clear in guidance that automated decision systems can produce unlawful discrimination even when nobody programmed an explicit preference for or against a protected group.

That route has a hard limit. The law addresses discrimination. It does not create a general entitlement to know why an application failed.

An applicant can file a complaint with the Division on Civil Rights or pursue a claim in court when the facts support one. The agency can investigate records and hiring practices. Litigation can eventually produce discovery, the formal exchange of evidence that may expose validation studies, scoring criteria or communications between an employer and a vendor. Those tools arrive after the applicant identifies a legal theory and enters an adversarial process.

The rejection notice itself does not have to contain the model’s reason.

This distinction works for employers. A person who knows only that software rejected a résumé may suspect disability screening, age discrimination or a proxy for race, but suspicion does not reveal which institution can compel records. The evidence needed to sharpen the complaint may be held by the employer, the software vendor or another data broker. Each can point toward the others.

New Jersey’s privacy law does not close the gap. The New Jersey Data Privacy Act, which took effect in January 2025, gives consumers rights over certain personal data and lets them opt out of some profiling, meaning automated processing used to predict behavior or make consequential decisions. Employment data sits outside much of that consumer framework. A job applicant cannot assume that a privacy-law access request will reveal an employer’s screening file.

Bills aimed specifically at automated employment tools have circulated in Trenton. A proposal is not a disclosure right. Until legislation is enacted and enforceable, it cannot make the automated notice explain itself.

Cross the Hudson and the paperwork changes

New York City’s Local Law 144 is the clearest nearby attempt to regulate automated employment decision tools, or AEDTs, software that substantially assists or replaces judgment in hiring or promotion. Enforcement began in July 2023.

Before using a covered tool, an employer must arrange an independent bias audit. The audit tests selection or scoring rates across specified race, ethnicity and sex categories, including intersectional categories, and the employer must publish a summary. Candidates must receive advance notice that the tool will be used, along with information about the qualifications or characteristics it will assess. The rules also address requests for information about data collection and retention.

That sounds like an answer. It is mostly an answer about the system, not the person.

A bias audit can show that one group advanced at a lower rate than another. It does not tell Mobley why a particular résumé was rejected, reproduce the score or require a human to reconsider it. Nor does a passing audit prove that the tool is fair. The published calculation is bounded by the employer’s historical data, the categories the rule specifies and the vendor’s definition of the tool being tested.

Coverage is bounded too. The law follows the employer’s use of an AEDT within New York City’s jurisdiction, not the applicant’s home address alone. A New Jersey resident applying to a covered New York City position may receive its protections. The same person applying to a job based elsewhere may not.

Software that assists a decision without meeting the rule’s threshold can also fall outside it.

Complaints about Local Law 144 go to the New York City Department of Consumer and Worker Protection. A claim that the resulting decision discriminated belongs elsewhere, potentially with the New York City Commission on Human Rights, the New York State Division of Human Rights or the federal Equal Employment Opportunity Commission. One tool. Several doors.

The city’s law makes the market more visible because employers must identify covered automation and post audit summaries. It still leaves the rejected applicant holding an automated notice that says little about the individual result.

Pennsylvania offers even less at the screening stage

Across the Delaware River, Pennsylvania has no comparable statewide audit-and-notice law for general automated hiring tools. The Pennsylvania Human Relations Act still prohibits employment discrimination, and Philadelphia has its own civil-rights enforcement structure, but those laws again require the applicant to frame the problem as discrimination rather than unexplained automation.

An employer may have internal appeal procedures. A vendor may describe its model in marketing material. Neither is the same as a statutory right to inspect the score, identify the decisive input or demand human review.

Maryland regulates one narrow practice: an employer’s use of facial-recognition technology during a pre-employment interview requires the applicant’s consent. That law reaches a visible biometric technique, not the quieter résumé-ranking systems that assign scores from work history, language patterns or imported assessments. It can stop a face scan from happening without permission. It does not explain the later rejection.

The regional map therefore changes according to the product feature. Facial analysis may trigger consent in Maryland. A covered ranking tool may trigger notice and an audit in New York City. An ordinary résumé screener in New Jersey or Pennsylvania may trigger no tool-specific disclosure at all.

The strongest notice may come from an older credit law

The Fair Credit Reporting Act can sometimes reach automated hiring systems, although its name makes the route easy to miss. The federal law covers reports supplied by a consumer reporting agency for employment purposes, including certain assessments of a person’s character, reputation or personal characteristics. The Federal Trade Commission has warned that calling a product artificial intelligence does not remove consumer-reporting obligations.

If a vendor and its product fall within that definition, an employer generally must obtain authorization before procuring the report. Before taking adverse action based on it, the employer must provide the applicant with a copy and a summary of rights. A later notice identifies the reporting agency, and the applicant can dispute inaccurate or incomplete information.

This is the closest thing in the patchwork to a practical paper trail. It can reveal the file used against an applicant and identify the company responsible for maintaining it. It still may not disclose the model’s weights or produce a plain-language account of why the score crossed a threshold.

The classification is also contested terrain. A vendor may argue that it supplies software rather than a consumer report. An employer may say the tool generated recommendations from information the applicant provided directly. The applicant, looking again at the rejection notice, usually cannot determine which description is legally correct from the outside.

Federal discrimination law supplies another route. Title VII covers race, color, religion, sex and national origin; the Age Discrimination in Employment Act covers workers 40 and older; the Americans with Disabilities Act addresses disability discrimination and inaccessible selection procedures. The EEOC can receive charges and investigate employers. These statutes can reach tests and algorithms that create unlawful disparate effects, but they are complaint systems, not automatic explanation systems.

Mobley’s litigation shows the price of that design. The applicant must assemble enough facts to allege discrimination before gaining access to the machinery that may prove or disprove it. The automated rejection remains the anchor, but the law asks the rejected person to identify the defect without seeing inside the tool.

Authority is the missing feature

The fragmented system is often described as a collection of protections. From the applicant’s side, it behaves like a routing exercise.

The New Jersey Division on Civil Rights handles discrimination under state law. The EEOC handles federal employment discrimination. New York City’s consumer agency enforces the city’s audit and notice rule. Human-rights agencies address discriminatory outcomes.

Consumer regulators and private disputes under the Fair Credit Reporting Act may become relevant when a vendor supplies a covered report. None serves as a general AI hiring ombudsman with authority to demand an individual explanation from every employer and vendor.

That fragmentation also determines what becomes public. New York City produces posted audit summaries. Discrimination complaints may produce findings, settlements or court records. Consumer-reporting disputes focus on the applicant’s file.

A rejection that is unfair, badly designed or impossible to understand but not provably discriminatory can pass between those systems without becoming any agency’s case.

An enforceable explanation right would have to name the responsible actor and specify the artifact an applicant receives: the data used, the principal factors behind the result, the vendor responsible, the route to human review and the regulator that can compel compliance. Nearby law divides those elements among different statutes. Employers buy one screening workflow. Applicants inherit the seams.

This article describes public laws and complaint structures, not legal advice. Deadlines, coverage and available claims depend on the facts and jurisdiction.

Questions people ask

Can

I force a New Jersey employer to explain an AI rejection?

Usually not under a general AI hiring law. New Jersey’s discrimination law may support an investigation or lawsuit when there is evidence of unlawful bias, while the Fair Credit Reporting Act may provide a report and dispute rights if a covered consumer reporting agency supplied the screening information.

Does

New York City’s bias-audit law apply to New Jersey residents?

Residency alone does not decide coverage. Local Law 144 focuses on covered use of an automated employment decision tool within New York City’s jurisdiction, so a New Jersey applicant for a covered city position may receive notice even though the same applicant would not receive it for every job.

Does a bias audit tell me why I was rejected?

No. The audit compares group selection or scoring rates and gives the public information about the tool’s aggregate results. It does not require the employer to disclose an applicant’s exact score, reveal every model input or prove that a particular rejection was free from discrimination.

Where can

I complain about an automated hiring tool?

The answer depends on the alleged violation. New Jersey discrimination complaints can go to the Division on Civil Rights, federal charges to the EEOC, and New York City audit-and-notice complaints to the Department of Consumer and Worker Protection. A consumer-reporting dispute follows a different route through the reporting agency and federal law.