Filter by Category

How to Challenge an AI Hiring Tool for Disparate Impact Under New York City Local Law 144 (AEDT Law) Explained

NYC Local Law 144 requires employers using automated employment decision tools (AEDTs) to complete an annual bias audit and provide candidate notice before use. For applicants and employees, that creates concrete hooks to challenge AI hiring systems that cause disparate impact—especially where audits, notices, or recordkeeping are deficient. This article explains how to evaluate an AEDT claim, preserve evidence, and pursue enforcement through NYC rules, administrative pathways, and complementary discrimination laws.

New York City’s Local Law 144 of 2021—commonly called the “AEDT Law”—is one of the first U.S. laws to directly regulate employers’ use of AI-driven hiring and promotion tools. While the law is framed as a transparency-and-audit statute rather than a private right of action, it can materially strengthen disparate impact challenges by forcing disclosures, creating compliance duties, and generating documents (bias audits, notices, postings) that can be leveraged in investigations and litigation.

This guide explains how attorneys can evaluate whether an AI hiring system is an “AEDT,” identify legal theories for disparate impact, and build a practical evidence plan. It also covers enforcement mechanics under NYC rules and how to pair Local Law 144 with federal, New York State, and New York City discrimination laws.

1) What Local Law 144 Requires (and Why It Matters for Disparate Impact)

Local Law 144 applies to employers and employment agencies that use an Automated Employment Decision Tool to screen candidates or employees for hiring or promotion within New York City. In broad strokes, the law requires:

Annual independent “bias audit” before use

An employer may not use an AEDT unless it has undergone a bias audit within the prior year. The audit is performed by an independent auditor and must evaluate the tool’s selection outcomes across protected categories using specified calculations. From a disparate impact perspective, this requirement can:

Create discoverable audit reports, datasets, and methodology documents.
Highlight adverse selection rates (or obscure them, depending on how the audit is constructed), providing a starting point for expert review.
Support negligence and willfulness arguments if the employer uses an unaudited tool or relies on an audit that is facially deficient.

Candidate/employee notice before the AEDT is used

Employers must provide notice that an AEDT will be used in the assessment or decision-making process. Proper notice is important because it can trigger earlier evidence preservation, enable targeted accommodation requests, and establish that the employer represented a tool as being used “objectively” or “fairly.” Failures or inconsistencies in notice can also undermine the employer’s credibility.

Public posting of audit summary information

Employers must post a summary of the most recent bias audit and the distribution date of the tool. For counsel, public postings can be collected immediately (before employers update or remove content) and used to frame pre-suit demand letters, administrative charges, and early motion practice.

Key limitation: Local Law 144 primarily authorizes enforcement by NYC (through civil penalties). It is not designed as a standalone damages statute for rejected applicants. But it can be a powerful evidentiary and regulatory “lever” in broader discrimination claims.

2) Is the Hiring System an “AEDT” Under the NYC Rules?

Whether the tool qualifies as an AEDT can determine whether Local Law 144’s audit and notice duties apply. In practice, tools commonly implicated include:

Resume screening and ranking algorithms
Chatbot “pre-screen” decision systems
Video interview scoring (e.g., automated scoring of speech patterns, word choice, facial movements, or prosody)
Game-based assessments producing automated fit scores
Automated “knockout” questionnaires with scoring thresholds that meaningfully shape who advances

Important nuance: many employers characterize the AI as merely “assistive.” For Local Law 144 analysis, counsel should focus on whether the system generates scores, rankings, classifications, or recommendations that are used to “substantially assist or replace” discretionary decision-making in hiring or promotion. Internal policies, vendor documentation, and recruiter workflows often contradict the marketing gloss.

3) What “Disparate Impact” Means in AI Hiring Cases

Disparate impact is a theory of discrimination that targets facially neutral practices that disproportionately exclude members of protected classes and are not justified by business necessity (or where less discriminatory alternatives exist). In AI hiring, disparate impact concerns arise because algorithms may:

Learn historical patterns reflecting past discriminatory practices
Use proxies for protected traits (e.g., zip codes, gaps in employment, school names, speech patterns)
Penalize non-standard communication styles or disability-related traits
Encode “cultural fit” in ways that systematically exclude groups

Local Law 144 does not replace Title VII, the New York State Human Rights Law (NYSHRL), or the New York City Human Rights Law (NYCHRL). Instead, it can supply key facts: when the tool was used, what it purported to measure, who audited it, what metrics the employer chose, and whether the employer followed required procedures.

4) Building a Claim: A Practical Attorney Checklist

A. Identify the employment action and decision chain

Start with the tangible outcome: rejection, failure to advance, lower ranking, or denial of promotion. Then map the decision chain:

Which steps were automated (screening, ranking, interview scoring)?
Who set thresholds and weights—employer or vendor?
Was there a human review stage, and did it meaningfully change outcomes?

B. Capture Local Law 144 compliance artifacts immediately

Before sending demands, preserve publicly available evidence:

Screenshots/PDFs of the employer’s bias audit posting and the “distribution date”
Job posting language referring to automated screening
Candidate notices (emails, portal banners, application screens)
Vendor product pages describing scoring, features, and data inputs

If you represent the candidate early, instruct them to download their application portal history and preserve timestamps, scoring outputs, and automated messages.

C. Look for “compliance gaps” that support inferences

Common Local Law 144 red flags include:

No public bias audit summary, or an audit dated more than 12 months prior to use
Audit summary that omits required selection rate information or uses unclear categories
Notice given after the assessment (or buried in a link that is not reasonably visible)
Inconsistent statements about whether AI was actually used

These issues may not automatically prove disparate impact, but they can support arguments for spoliation prevention, adverse inferences, and the need for expedited discovery—particularly if the employer is continuing to use the tool.

5) How to Use Bias Audits in a Disparate Impact Strategy

Bias audits under Local Law 144 typically focus on selection rate comparisons across categories (for example, sex, race/ethnicity). In disparate impact litigation, your expert will often need more detail than the posted summary provides. Still, the audit can shape the case in several ways:

Use the audit to define the “practice” at issue

Disparate impact claims require specifying the challenged practice. The audit may identify the exact tool, model version, and how it is used (e.g., “automated ranking of applicants for interview selection”). That specificity helps withstand motions attacking pleading sufficiency.

Spot methodological weaknesses

Audits may be limited by:

Small sample sizes that mask disparities
Overbroad job groupings that dilute impact
Using “applicant flow” data that excludes individuals deterred by the system
Failure to evaluate intersectional effects (e.g., Black women) where permitted by other laws even if the audit summary is more general

Probe what the audit did not test

Many AI tools implicate disability, age, and language-related impacts. If the audit summary focuses narrowly on a subset of categories, counsel should explore whether the tool’s features (video scoring, typing speed tests, gamified cognition measures) could create disability-related disparate impact or require accommodation under disability law.

6) Examples of Fact Patterns That Can Support a Challenge

Example 1: Video interview scoring penalizes speech patterns

A candidate completes an automated video interview. The platform provides a “communication score” that correlates with accent, speech cadence, or disability-related traits. The candidate is rejected without human review. If the employer’s Local Law 144 notice is missing or vague, counsel can argue the candidate was deprived of a chance to request accommodation and that the employer concealed a selection practice that had predictable disparate impact.

Example 2: Resume screener uses proxies tied to race or socioeconomic status

An employer uses an AEDT to rank resumes and only interviews the top 10%. Inputs include school name, zip code, or work history features that proxy race or class. Disparate impact may be shown through applicant-flow analysis, while Local Law 144 postings and audit documents help establish the tool’s role and whether it was properly evaluated.

Example 3: “Knockout” assessments screen out caregivers

A tool scores “availability” or schedule flexibility and automatically rejects candidates who indicate constraints. The impact may fall disproportionately on women and caregivers. If the AEDT is used as a gatekeeper and the audit ignores the effect of certain questions, plaintiffs may argue that less discriminatory alternatives exist (e.g., allowing later scheduling discussions after assessing core qualifications).

7) Procedural Pathways: Enforcement and Complementary Claims

Local Law 144 enforcement

Local Law 144 is primarily enforced by NYC through civil penalties for noncompliance (e.g., using an AEDT without the required audit or failing to provide notice/posting). Practically, this means: