NYC Local Law 144: Bias Audit, Notice, and Penalties

NYC Local Law 144 requires employers and employment agencies to run an independent bias audit on any automated tool they use to screen job candidates or evaluate employees for promotion, publish a summary of the audit results, and notify affected candidates at least ten business days before the tool is used. The Department of Consumer and Worker Protection (DCWP) has enforced the law since July 5, 2023, and it remains the first municipal rule in the United States aimed specifically at algorithmic hiring tools.¹

Who the Law Covers

The trigger is where the job is, not where the applicant lives. If the position sits in New York City, the law applies even if the applicant is elsewhere. If the position sits outside the city, the law does not apply even when NYC residents apply for it.

Fully remote roles are the harder call. If the employer’s only office is in NYC, the law applies. If the employer has no NYC office, it does not. Employers with offices both inside and outside the city need to look at each remote role on its own, based on where the employee will actually work and which office they report to.

What Counts as an Automated Employment Decision Tool

An automated employment decision tool, or AEDT, is any computational process using machine learning, statistical modeling, data analytics, or artificial intelligence that produces a simplified output (a score, classification, or recommendation) and that substantially assists or replaces human discretion in hiring or promotion decisions.¹ DCWP’s guidance describes the underlying techniques as “mathematical, computer-based techniques” that generate a prediction or classification and also identify the inputs and their relative importance to improve accuracy.²

In practice, that captures resume-screening software that ranks applicants, video interview platforms that score verbal responses, and chatbot assessments that filter candidates, as long as the output feeds into a hiring or promotion decision. A “prediction” includes an assessment of fit or likelihood of success; a “classification” includes assigning a candidate to a tier or category.

The law explicitly excludes tools that do not automate or materially influence discretionary decisions. Spam filters, firewalls, antivirus software, calculators, spreadsheets, and basic databases are not covered. The dividing line is whether the tool’s output substantially shapes or replaces a manager’s choice. A spreadsheet that sorts applicants alphabetically is not an AEDT. Software that assigns a “fit score” and auto-rejects anyone below a threshold is.

The Bias Audit

Before an employer can use an AEDT, the tool must have gone through an independent bias audit conducted no more than one year earlier. The audit tests whether the tool produces different selection or scoring rates across demographic groups, using the same job categories employers already report to the federal EEOC (the EEO-1 component categories).¹

At a minimum, the auditor has to calculate selection or scoring rates and their corresponding impact ratios across three dimensions: sex categories, race and ethnicity categories, and intersectional categories that combine sex with race or ethnicity.² The intersectional piece is where most audits get complicated. Showing that men and women are treated equally overall, and that racial groups are treated equally overall, is not enough. The audit also has to check whether groups like Black women or Hispanic men are selected at disproportionately low rates compared to the most-favored intersectional group.

An impact ratio compares the selection rate of one group to the selection rate of the most-selected group. If a category represents less than 2% of the audit data, it can be excluded from the required calculations, but every other category has to be included. The auditor works from historical data collected during the employer’s actual use of the AEDT, so employers must hand that data over.

The auditor has to be independent. That means they cannot have been involved in developing the AEDT or in using it to make employment decisions. The law does not spell out credentials or certifications for auditors. Because a new audit is required every year, this is a recurring operational cost.

One thing the audit does not do is protect the employer from discrimination claims. It measures statistical disparities in the tool’s output. It does not evaluate whether the underlying model is valid, whether the training data was representative, or whether the job qualifications the tool screens for are lawful. A clean audit is compatible with real liability under federal and state anti-discrimination law.

What You Have to Publish

The employer or agency has to post a summary of the most recent bias audit on its website before using the tool. The summary needs a clear and conspicuous link and has to stay posted for at least six months after the AEDT was last used for an employment decision.¹

The summary must include the date of the audit, the selection or scoring rates and impact ratios for every required category (sex, race and ethnicity, and intersectional), the number of applicants or candidates in each demographic category, and the source and explanation of the data used to conduct the audit.²

Published summaries have been a common trouble spot. A New York State Comptroller review released in 2025 found at least 17 potential compliance problems among 32 companies it examined (missing impact ratios, incomplete demographic breakdowns, or summaries that omitted required information), even though DCWP’s own review of the same companies flagged only one issue.³ An absence of enforcement action is not evidence that a disclosure is compliant.

Notice to Candidates

Every candidate or employee who lives in NYC has to be notified at least ten business days before an AEDT is used to evaluate them. The notice can go by email, by postal mail, or in the job posting itself.¹

Two things have to be covered. First, the notice must tell the person that an AEDT will be used in their assessment and explain how to request an alternative selection process or a reasonable accommodation under other applicable laws. Second, it must identify the specific job qualifications and characteristics the tool will evaluate.

Including those instructions does not obligate the employer to actually provide an alternative selection process. DCWP’s rules state that “nothing in this subchapter requires an employer or employment agency to provide an alternative selection process.”⁴ The notice is an information requirement, not a guarantee of human-only review. Whether a reasonable accommodation is available is a question for other laws, such as the Americans with Disabilities Act or the NYC Human Rights Law.

Data Retention and Source Disclosures

Separately, the employment section of the employer’s website has to describe the AEDT data retention policy, the type of data the tool collects, and the source of that data.⁵ The site must also include instructions for submitting a written request for this information, and the employer has 30 days to respond. If disclosure would violate another law or interfere with a law enforcement investigation, the employer has to explain why.

Penalties

DCWP can impose civil penalties for violations, and the structure is per-violation and per-day:

Up to $500 for the initial violation on the first day.
Between $500 and $1,500 for each additional violation after the first day.¹

Each day of non-compliance counts as a separate violation for bias audit failures. Each candidate who does not receive the required notice counts as a separate violation. An employer running a large hiring campaign without proper notices could rack up hundreds of individual violations in a matter of weeks. The law does not give candidates a private right of action, so they cannot sue in court under Local Law 144 itself, but they can file administrative complaints with DCWP.

What Enforcement Actually Looks Like

Enforcement has been thin. The Comptroller’s audit covering July 2023 through June 2025 found that DCWP received only two AEDT-related complaints in the law’s first two years, and the agency did not examine whether its complaint intake process was actually functioning.³ The takeaway for employers is practical: the enforcement gap can close at any time, and the per-day, per-candidate math means catching up later could be expensive.

Federal Law Still Governs the Substance

Local Law 144 is a transparency and process rule. It requires audits, disclosures, and notices, but it does not set a threshold for how much disparate impact is too much. Federal anti-discrimination law does that work. The EEOC has confirmed that Title VII, the Age Discrimination in Employment Act, the ADA, and related statutes apply to AI-driven hiring tools the same way they apply to any other employment practice.⁶

A tool that screens out candidates in a pattern correlated with race, sex, age, disability, or another protected characteristic can trigger a disparate impact claim even without any intent to discriminate. The employer’s defense in that scenario requires showing that the criteria are job-related and consistent with business necessity. A passing Local Law 144 audit does not establish that defense. The audit only shows whether different selection rates exist, not whether those differences are legally justified. Compliance in NYC is dual: procedural under Local Law 144, and substantive under federal and state anti-discrimination law.

1
NYC Department of Consumer and Worker Protection. Automated Employment Decision Tools (AEDT)
2
NYC Department of Consumer and Worker Protection. Automated Employment Decision Tools: Frequently Asked Questions
3
Office of the New York State Comptroller. Enforcement of Local Law 144 – Automated Employment Decision Tools
4
NYC Rules. Automated Employment Decision Tools
5
NYC Department of Consumer and Worker Protection. New York City Department of Consumer and Worker Protection
6
U.S. Equal Employment Opportunity Commission. What is the EEOCs Role in AI