If your company uses any AI or algorithmic tool to help filter, score, or rank job candidates — even a commercial applicant tracking system with a “best match” feature — you may already be subject to New York City’s Local Law 144. Most hiring managers have heard the name. Fewer understand what it actually requires, what triggers it, and what using a covered tool without compliance means in practice.
This piece is written for technical hiring managers: people who use AI to screen candidates for roles where real skill differences matter — engineers, estimators, project managers, technical specialists. The goal is an accurate account of what the law requires. This is not legal advice and should not substitute for a review by employment counsel familiar with your specific situation and current enforcement activity in your jurisdiction.
What Local Law 144 Is
New York City Local Law 144 of 2021 applies to employers and employment agencies operating in New York City that use “automated employment decision tools” (AEDTs) in hiring or promotion decisions affecting New York City candidates or employees. The law took effect in 2023, with enforcement by the Department of Consumer and Worker Protection (DCWP) following implementation rules published in the months after. Enforcement details and rule interpretations have continued to develop; verify current requirements with counsel.
The law was designed to address a specific risk: an employer who deploys an AI screening tool has no independent check on whether that tool produces outcomes that discriminate against candidates based on race, sex, or other protected characteristics. An algorithm trained on historical hiring data can reproduce the patterns of that history, including its biases, without anyone intending that result. The law’s answer is transparency and third-party audit — before you use the tool on your candidates, have someone independent check it.
What Counts as an AEDT
The law defines an automated employment decision tool as any computational process — derived from machine learning, statistical modeling, data analytics, or artificial intelligence — that issues a simplified output (a score, classification, or recommendation) used to “substantially assist or replace” discretionary decision-making for hiring or promotion.
The phrase “substantially assist or replace” is where most of the practical ambiguity lives. A tool that produces a pass/fail score that filters your candidate pool before any human reviews rejected applications likely qualifies. A hiring manager who uses an AI model to help think through a résumé — reads the AI’s output, then makes a holistic human judgment that they could articulate independently — is in a different position. The line between assistance and substitution is not always clean, and where your use case falls on that line is a determination to make with counsel, not an inference to guess at.
Likely examples of covered tools: commercial ATS platforms with algorithmic matching scores that route candidates into interview or reject queues without human review of the excluded set; résumé screening tools that classify candidates as qualified or unqualified; video interview analysis tools that score expression, tone, or language.
Less likely to be covered: using a general-purpose AI model to help draft interview questions, write a job description, or prepare for a debrief — where the output does not directly assess whether a specific candidate advances.
The Three Requirements
If you are using a covered AEDT for NYC hires or promotions, Local Law 144 requires three things.
An independent bias audit, conducted annually. Before using the tool for any covered decision, and at least once per year after that, an independent auditor — someone without a financial interest in the tool’s deployment — must conduct a bias audit. The audit calculates selection rates and impact ratios: how the tool’s outputs differ across categories of sex and race/ethnicity. The employer is responsible for obtaining this audit, not the vendor.
Some commercial vendors provide bias audit documentation for their tools. Before relying on it, confirm whether the audit covers your specific deployment — your candidate population, your use case, your data — and not only the vendor’s baseline performance on generic test data. A vendor audit of their tool’s average behavior is not the same as an audit of how the tool performs on your actual candidates.
Notice to candidates. Candidates must be notified that an AEDT will be used in their hiring process. Under the law’s requirements, this notice must be given at least 10 business days before the tool is applied to their candidacy. The notice must describe what characteristics the tool evaluates and what data it collects. Candidates must be given the opportunity to request an alternative selection process or a reasonable accommodation.
This has real operational implications. If you are running applications through a screening tool within the first day or two after submission, you need a notice mechanism built into your application flow — not buried in a general privacy policy.
Publication of the most recent audit summary. The bias audit summary must be publicly available — on the employer’s website or via a link in the job posting. The published information must include the date of the audit, the date the AEDT was first used, and a description of the data used to train and test the tool.
Penalties and Enforcement
The DCWP assesses civil penalties for violations on a per-violation, per-day basis. The specific penalty range is established by the agency’s rules; confirm current amounts with counsel, as enforcement guidance has continued to develop since the law’s implementation. This is a compliance area where the regulatory picture is active — not a set-and-forget determination you made in 2023.
The EEOC Angle: National, Not Just NYC
Title VII of the Civil Rights Act of 1964 prohibits employment practices that have a disparate impact on protected classes, even when there is no discriminatory intent. The EEOC has made clear that this applies to AI and algorithmic tools used in employment selection. In May 2023, the agency published technical assistance guidance specifically on assessing adverse impact in software, algorithms, and AI used in hiring.
The practical implication is significant: even if your company is not based in New York City and your candidates are not New York City residents, if your screening tool produces outcomes that disproportionately screen out candidates in a protected class, you may face EEOC exposure. The tool’s vendor did not intend to discriminate. You may not have intended to discriminate. Intent is not the standard under disparate impact doctrine.
Bias auditing matters beyond the NYC law for exactly this reason — it is the mechanism that gives you documented evidence about whether disparate impact is occurring and a record that you checked.
The EU AI Act: If You Hire Globally
The EU AI Act (Regulation 2024/1689) entered into force in August 2024 with phased implementation. It classifies AI systems used in employment — including recruitment, candidate selection, and employee management — as high-risk systems subject to stricter requirements: technical documentation, human oversight requirements, accuracy and robustness standards, and transparency to the individuals the system affects.
If your company hires candidates in EU member states, or if your AI screening tools come from vendors operating under EU jurisdiction, confirm with EU employment counsel whether the Act’s requirements apply to your workflows and under which implementation timeline. The compliance obligations are layered and continue to phase in.
What to Do If You Are Already Using AI Screening
If you are currently using any algorithmic or AI-based tool to score, rank, or filter candidates, four steps are worth taking before your next hire.
First, ask your vendor directly: Does this tool produce simplified outputs used to influence hiring decisions? Does the vendor provide bias audit documentation for your specific deployment, not just their generic baseline? What is their stated position on Local Law 144 compliance and EEOC guidance?
Second, confirm where your candidates are located. If you are hiring for roles in New York City, determine whether your process triggers Local Law 144 and, if so, whether you have met the notice, audit, and publication requirements before using the tool.
Third, document the role the AI plays in your process. A workflow where AI filters the candidate pool and rejected candidates are never reviewed by a human is materially different from one where AI surfaces candidates and humans review the full picture, including who was excluded. The documentation protects you if a hiring decision is ever challenged.
Fourth, keep a human in the actual decision. The legal risk — and the reputational risk — concentrates when an AI output makes or substantially drives a final call without a human who understands and can articulate the basis for that call. A tool that assists a human decision-maker who reviews the underlying evidence is in a fundamentally different legal and ethical position than a tool whose output determines who advances without human review.
The Architecture Behind This Advisory
Grounded Hire is an AI advisory for hiring managers, written from the perspective of AI architecture — not HR and not law. The work here focuses on a specific problem: AI tools in hiring that produce outputs without showing their reasoning, citing their sources, or acknowledging their limits create legal exposure for the employers who rely on them and real harm for the candidates they assess.
The alternative is AI assistance that cites what it found in the résumé and job description, distinguishes what it can determine from what it cannot, and keeps a human who understands the basis of every decision in the loop. That is not a weaker tool. It is a more defensible one — and one that can actually be audited.