Racial Discrimination In AI Screening

You upload a resume, answer a few questions, maybe take a timed assessment, and click submit. The portal refreshes. Sometimes the rejection comes in minutes. Sometimes it’s silence. Or it’s a polite note saying the company “went in another direction,” even though the job stays posted and you meet the requirements.

For many applicants, that’s “just modern hiring”. For applicants of color, it can feel different — like an invisible filter, a decision you can’t question, and a system that keeps saying no without ever explaining why.

That’s the risk of poorly designed AI screening. It can automate old inequities, scale them, and then hide them behind the language of objectivity. New Jersey made this explicit: the anti-discrimination law applies to algorithmic bias the same way it applies to bias committed by real people.

At Brandon J. Broderick, we see how these cases surface in the real world — not as technical debates about software, but as patterns of exclusion that are hard to see and harder to challenge. 

Let’s break down how racial bias enters AI screening, how “neutral” criteria can still exclude, why these systems are so difficult to challenge, and when it’s time to talk to a racial discrimination lawyer in New Jersey.

The Legal Basics: Federal And New Jersey Law Still Apply To Algorithmic Bias

AI tools don’t exist outside the law. They operate within rules that already govern hiring decisions. Federal law hasn’t changed simply because technology has.

Under Title VII of the Civil Rights Act, employers can violate the law in two main ways:

• Intentional discrimination — using race, sex, or other protected traits directly
• Disparate impact — using facially neutral hiring tools that disproportionately exclude protected groups without a valid business justification

The Equal Employment Opportunity Commission has made this clear: AI screening tools are treated like traditional hiring tests when they function as selection procedures.

That means:

• If an AI tool screens resumes, ranks candidates, or filters applicants, it can trigger disparate-impact analysis
• Employers are expected to evaluate if the tool produces significantly different selection rates for protected groups
• The same testing standards used for decades still apply

In 2024, the EEOC filed an amicus brief in the Workday AI screening litigation, arguing that software vendors themselves may qualify as “employment agencies” under Title VII when their tools perform large-scale screening functions.

Recent Development In New Jersey: NJLAD And Algorithmic Bias

New Jersey’s main civil rights law is the New Jersey Law Against Discrimination (NJLAD), enforced by the New Jersey Division on Civil Rights within the New Jersey Office of the Attorney General.

In January 2025, the Attorney General and DCR issued formal guidance stating plainly that NJLAD applies to algorithmic discrimination — meaning employers can be liable when automated hiring or screening tools produce unlawful biased outcomes.

A recent survey found that New Jersey employers are rapidly adopting new technology:

• 63% of employers reported using at least one AI-enabled tool to recruit applicants or make decisions.
• At the same time, 47% of employers acknowledged that using AI may lead to unfair or biased outcomes.

This is why implicit bias training alone now falls short. Training managers and HR staff do little when an automated system filters applicants out before any human review. When that happens, consulting a racial discrimination attorney in New Jersey may be one of the few ways to meaningfully challenge what would otherwise remain hidden behind technology.

Why Algorithmic Bias In New Jersey Often Begins With Job Requirements, Not Code

AI screening doesn’t usually create bias on its own. It reflects whatever an employer defines as “qualified.”

Most tools simply translate job requirements and preferences into filters and rankings. That means bias can be built in before the system ever scores a candidate.

So-called “neutral” requirements are often the real problem. Common examples include:

• Degree inflation — jobs that used to require skills now require degrees, reflecting unequal access to education
• Prestige signals — preference for certain schools, employers, or titles that mirror historic opportunity gaps
• Resume “polish” rules — language, formatting, or style that tracks class and cultural norms, not ability
• “Culture fit” proxies — traits framed as teamwork or attitude that really reward similarity to existing staff

Language expectations themselves can amplify the problem. When employers unevenly enforce language policies or communication standards, such as penalizing accented speech or non-standard phrasing, and then encode those expectations into hiring criteria, the AI inherits that bias. What began as selective human judgment becomes a rigid automated filter.

When these preferences are built into an AI tool, the system can exclude candidates by race without ever using race as a factor. It’s not inventing discrimination — it’s automating it.

Training Data That Recycles Old Inequality In AI Hiring Under New Jersey Law

Hiring data reflects who historically had access to opportunities and who was excluded. That includes:

• who received internships, mentorship, and promotions
• who was labeled “high potential”
• who was pushed out of hostile workplace cultures
• who had resume gaps tied to caregiving, health, or economic barriers

In customer-facing roles, this biased history often includes another layer: “image” discrimination. Past hiring decisions may have favored candidates perceived as the “right fit” for customers — a vague standard that can mask discrimination based on race, accent, age, disability, or appearance.

When an AI system is trained to predict “successful employees” using this kind of data, it can treat exclusion as evidence of merit. The model doesn’t need to reference protected traits directly. A skewed history is enough.

Vendor Black Boxes And Outsourced Responsibility For Algorithmic Bias

One reason AI hiring bias is so hard to challenge is the vendor “black box.”

In many of the largest discrimination matters our team sees, the same pattern shows up early: the employer doesn’t really know how its own system works. 

Resume screeners, matching algorithms, automated assessments, rankings, chatbots, background scores — they’re bolted together and treated like utilities. Compliance is assumed to live somewhere upstream, inside the vendor contract.

When harm surfaces, the explanations are familiar:

• “We didn’t design it.”
• “We can’t see inside it.”
• “The vendor certified compliance.”
• “We follow the rankings.”

This is not new. It mirrors how some employers have historically defended racially skewed security practices — claiming they were simply following “protocols” when certain employees were stopped, monitored, or questioned more than others. The system becomes the excuse.

But outsourcing decision-making does not outsource responsibility. If an employer uses an automated tool and that tool produces an unlawful result, liability follows the outcome — not the code’s author.

What makes these cases especially striking is how often the black box becomes a shield. The opacity doesn’t protect workers. It protects distance. It allows discrimination to appear as a software glitch rather than a business decision.

And in practice, once you start pulling on the thread — asking who selected the tool, what inputs were chosen, what outcomes were reviewed, and what warnings were ignored — the “we didn’t know” story tends to collapse. 

How Automated Hiring Is Changing Discrimination Claims In New Jersey

Traditional discrimination cases often involved concrete evidence: a biased comment, uneven standards, shifting explanations, or a manager whose decisions left a paper trail.

Automated systems remove many of those signals. Instead of a person, there’s a system. Instead of an explanation, there’s silence. Common obstacles show up again and again:

• no clear decision-maker — the rejection feels automatic
• no reason given — applicants aren’t told what failed
• invisible comparisons — applicants can’t see who ranked higher or why

Discrimination doesn’t disappear; it simply becomes harder to see. That means proof often shifts from individual moments to patterns, statistics, and internal records — evidence most applicants can’t access without legal help.

This is why cases like the Workday litigation matter. The EEOC has argued that when algorithms make decisions at scale, civil rights laws still apply.

Race And Disability Intersection: When Screening Penalizes “Non-Standard” Lives

Many applicants of color may also live with disabilities, including anxiety, ADHD, depression, chronic illness, hearing loss, or conditions that affect how work histories look on paper. AI screening tools often penalize those realities while claiming neutrality.

That can show up in subtle ways:

• resume gaps treated as risk, not context
• non-linear careers flagged as instability
• timed assessments that ignore accommodation needs
• communication styles scored as “lower fit”
• past medical leave folded into reliability predictions

What our advisors and specialists tend to notice — across very different industries — is how often these “neutral” flags cluster around the same people. The same resume gap, the same job change, the same assessment result can yield very different outcomes depending on whose history it represents and what assumptions the system has been trained to make.

What gets labeled “non-standard” is rarely random. It reflects who had access to healthcare, flexibility, mentorship, and second chances — and who didn’t. Our team often sees how those structural gaps are quietly translated into scoring penalties long before any human ever reviews an application.

When an AI system treats “not perfectly linear” as “not qualified,” it predictably filters out people who already have to work around barriers. And those barriers don’t fall evenly across race or disability — even when the software insists it isn’t “looking” at either.

How AI Hiring In New Jersey Can Reinforce Racial Stratification Across Job Tiers

One of the most overlooked harms of automatic screening is who gets hired into which jobs. AI tools can quietly sort applicants into different job tiers:

• applicants of color screened out of higher-pay, higher-mobility roles
• applicants of color routed into lower-paid or high-surveillance positions
• “fit” scoring that mirrors historically segregated career tracks

That same sorting logic often carries into project assignments after hiring. Workers who score lower on opaque “fit” or “potential” metrics may be steered away from visible projects, leadership tracks, or revenue-driving assignments — while others get the work that leads to promotion and influence.

This isn’t usually explicit. It happens through design choices.

For example, lower-wage or frontline roles are more likely to use high-volume, opaque AI screens. Higher-level roles often involve smaller applicant pools, referrals, and greater reliance on human judgment. When AI is concentrated at the bottom of the job ladder, the pipeline itself starts to shape who moves up — and who never gets the chance.

Over time, this can change a company’s workforce without anyone ever saying the quiet part out loud. No single decision looks discriminatory. The system does the sorting.

What Accountability Really Means Under New Jersey Law

Accountability doesn’t require a perfect system — and it doesn’t require turning hiring into a science project. It only requires deliberate choices.

Employers can act now to determine whether AI is a helpful tool or a shield against accountability. 

Real accountability means testing outcomes for bias, making sure tools measure job-related skills rather than proxies for race or class, reviewing results over time, demanding transparency from vendors, and allowing meaningful human review before candidates are quietly filtered out.

If you believe an automated system cost you a fair chance, or you’re seeing patterns that don’t add up — you don’t have to navigate that alone.

Contact us for a free consultation to understand your rights under New Jersey and federal law.