AI Hiring Tools and Discrimination Risk

AI hiring-tool compliance depends on the tool, the data it uses, and where you operate. This post is informational. For your specific hiring process, consult a licensed attorney.

AI has moved into hiring quickly and quietly. Resume screeners rank applicants, chatbots conduct first-round interviews, video tools score candidates on tone and word choice, and automated systems decide who advances before a human sees a single application. The pitch is efficiency and objectivity. A machine, the reasoning goes, cannot be biased the way a person can.

The reasoning is wrong, and the legal exposure is real. An AI hiring tool can produce discriminatory outcomes even when no one involved intended to discriminate. And when it does, the liability generally lands on the employer that used the tool, not the vendor that built it.

Disparate impact: the core risk

Federal employment law does not only prohibit intentional discrimination. Under Title VII, a neutral practice that disproportionately screens out a protected group can be unlawful (the doctrine of disparate impact) unless the employer can show the practice is job-related and consistent with business necessity.

AI hiring tools are disparate-impact machines by design. They learn from historical data, and if that data reflects past patterns of who got hired, the model can reproduce and amplify those patterns. A resume screener trained on a workforce that skewed one way can learn to prefer candidates who resemble it, using proxies (a ZIP code, a gap in employment, a particular phrasing) that correlate with protected characteristics without ever naming them. The Equal Employment Opportunity Commission has made clear that using an automated tool does not shift this responsibility away from the employer.

FCRA when the tool uses background or consumer data

If an AI hiring tool draws on background checks or other consumer-report data supplied by a third party, the Fair Credit Reporting Act comes into play. The FCRA imposes specific obligations: disclosure to the candidate, authorization, and a defined adverse-action process if you decide not to hire based on that information. Automating the decision does not remove those steps; it just makes it easier to skip them and create liability.

The new wave of automated-decision-tool laws

On top of the established federal framework, a growing set of state and local laws now regulate automated employment decision tools directly. These laws vary, but the recurring themes are bias audits, candidate notice, and in some cases the right to request an alternative process. If you hire across multiple jurisdictions, you may be subject to more than one of these regimes at once, and they are still multiplying.

What to do before you deploy

The goal is to capture the efficiency of these tools without inheriting a discrimination claim. A few steps do most of the work:

Diligence the vendor, but do not rely on it. Ask whether the tool has been tested for adverse impact, on what data, and how recently. Get it in writing. But remember that the vendor’s assurances do not transfer your legal responsibility. You still own the outcome.

Test outcomes on your own data. The only way to know whether a tool is screening out protected groups in your applicant pool is to look. Periodic adverse-impact analysis of the tool’s actual results is the single most useful control.

Keep a human in the loop. A tool that ranks or flags candidates for human review is easier to defend than one that auto-rejects. Meaningful human judgment at the decision point matters, legally and practically.

Handle notice and consent correctly. Where the tool triggers FCRA obligations or a state notice requirement, build those disclosures into the process rather than bolting them on later.

Write down what the tool does. If a claim arises, you will need to show what the tool considered, how you vetted it, and what oversight you applied.

The bottom line

AI hiring tools are not off-limits, and used carefully they can genuinely improve a hiring process. But “the software decided” is not a defense. The employer chose the tool, deployed it, and acted on its output, and that is where the law places responsibility.

If you are using or considering AI in hiring, talk to MB Legal about the compliance and diligence that keep an efficiency gain from becoming a liability.

This post is informational and does not constitute legal advice. Employment-discrimination and AI hiring-tool obligations vary by jurisdiction and fact pattern. For guidance specific to your hiring process, consult a licensed attorney.