NYC Local Law 144: AI Hiring Bias Audits Explained
NYC Local Law 144 is a New York City statute that bars employers and employment agencies from using an automated employment decision tool (AEDT) to screen candidates or employees for a position in NYC unless the tool has passed an independent bias audit within the prior year, the results are published publicly, and affected candidates receive advance notice. Enforcement began July 5, 2023, and violations carry civil penalties of up to $500 for a first violation and between $500 and $1,500 for each subsequent or continuing violation, with each day of non-compliant use counted separately.
The law applies to hiring and promotion decisions for jobs located in New York City, including remote roles tied to a NYC office. It is among the first U.S. mandates tying the use of hiring algorithms to a published, third-party statistical audit.
When did Local Law 144 take effect, and why did enforcement slip?
Local Law 144 was enacted in December 2021 but did not take effect until January 1, 2023, and the DCWP did not begin enforcing it until July 5, 2023. The gap was driven by a rulemaking process that ran longer than the statute anticipated. The DCWP issued proposed rules in September 2022 to interpret the statute's open terms, drew a large volume of public comment, withdrew the first draft, and issued a second set of proposed rules in December 2022. The deferred enforcement date explains why employers running AEDTs in early 2023 were subject to the law before any audit method was final.
Milestone: Local Law 144 enacted.
Date: December 2021.
Milestone: First proposed DCWP rules.
Date: September 2022.
Milestone: Second proposed DCWP rules.
Date: December 2022.
Milestone: Law takes effect.
Date: January 1, 2023.
Milestone: Enforcement begins.
Date: July 5, 2023.
What is an automated employment decision tool (AEDT) under Local Law 144?
An automated employment decision tool is any computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence that issues a simplified output (a score, classification, or ranking) used to substantially assist or replace discretionary human decision-making in hiring or promotion.
The "substantially assist or replace" condition is the operative trigger. Under the Department of Consumer and Worker Protection (DCWP) final rules, a tool falls under the law when it does one of the following:
Relies solely on a simplified output with no other factors considered.
Weights a simplified output more than any other criterion in the decision set.
Uses a simplified output to overrule conclusions derived from other factors, including human judgment.
Tools whose outputs a human reviewer weighs equally alongside many independent factors generally fall outside the trigger. Because this distinction decides whether screening software is regulated or exempt, employers should document how each tool feeds the final decision.
Which tools are covered and which are exempt?
Covered tools include resume scorers, video-interview analyzers, ranking engines, and matching algorithms that drive candidate selection.
Tool or use case: Resume-ranking algorithm that orders applicants by score.
Covered by Local Law 144? Yes.
Reason: It produces a simplified output that directly influences hiring or screening decisions.
Tool or use case: Chatbot that schedules interviews only.
Covered by Local Law 144? No.
Reason: It does not generate a simplified output used to assist a hiring decision.
Tool or use case: Assessment that automatically rejects candidates below a predefined cutoff.
Covered by Local Law 144? Yes.
Reason: The automated output directly determines candidate eligibility by overriding other factors.
Tool or use case: AI tool used for a job located entirely outside New York City.
Covered by Local Law 144? No.
Reason: The position falls outside the geographic scope of Local Law 144.
Tool or use case: Software that a recruiter consults as one of several equally weighted inputs.
Covered by Local Law 144? Generally no.
Reason: The AI output is not weighted more heavily than other factors in the hiring decision.
What does a Local Law 144 bias audit require?
A bias audit is an impartial evaluation by an independent auditor who has not used, developed, or distributed the tool and has no financial interest in the employer or vendor. The audit must calculate selection or scoring rates across demographic categories and compute impact ratios that measure disparate outcomes.
The required calculations are:
Selection rate for each sex category, race/ethnicity category, and intersectional category (sex combined with race/ethnicity).
Impact ratio, defined as the selection rate of a group divided by the selection rate of the most-selected group, for each category.
Scoring rate analysis for tools that output a continuous score rather than a binary pass/fail, comparing the rate at which each group scores above the sample median.
Which demographic categories does the audit use?
The race/ethnicity and sex categories follow the EEO-1 Component 1 classifications, the same framework employers use for federal workforce reporting. The sex categories are male and female. The race and ethnicity categories are Hispanic or Latino, White, Black or African American, Native Hawaiian or Other Pacific Islander, Asian, American Indian or Alaska Native, and two or more races. The intersectional analysis crosses each sex category with each race/ethnicity category, producing more than two dozen subgroups in a full audit. The audit reports the applicant count in each category and may exclude any category below 2% of the data, noting that exclusion in the published results.
What data can an auditor use?
Auditors must use historical data from the employer's actual use of the tool. When an employer has insufficient history, the rules permit test data, but the published summary must state why historical data was not used. Multiple employers using the same vendor tool may rely on a single audit if they contribute their own historical data or the audit otherwise reflects their use, which is how most vendor-level audits operate.
What makes the data hard to collect in practice?
The audit depends on demographic data the law does not let employers compel. Race, ethnicity, and sex are self-reported by candidates, usually through a voluntary equal employment opportunity question during the application. Many applicants decline to answer, so a meaningful share of records carry no demographic label and cannot be placed in any category, shrinking the population an auditor can analyze. The under-2% exclusion compounds this by dropping the smallest groups, and a newly adopted tool may have no usage history at all, forcing reliance on test data that reflects the vendor's sample rather than the employer's own applicant pool.
Does the vendor's audit cover the employer?
Most AEDTs are licensed from vendors, and most published audits are commissioned by those vendors rather than by individual employers. A shared vendor-level audit can satisfy the law for many clients at once, but only if it reflects how each employer actually uses the tool. The legal obligation still sits with the employer or employment agency. Before deploying a tool in NYC on a vendor audit, an employer should confirm the audit is less than one year old, covers the same version and configuration the employer runs, and either reflects the employer's own historical data or otherwise represents its use. If the employer's use diverges from the audited use, a vendor's passing result may not cover it.
How do employers publish results and notify candidates?
Beyond commissioning the audit, compliance carries two public-facing duties.
Publication. Employers must post a summary of the most recent bias audit on the careers or jobs section of their website in a clearly visible spot. The summary must include the audit's distribution date, the source and explanation of the data, the selection or scoring rates, and the impact ratios for all required categories. The audit cannot be more than one year old.
Notice. Employers must notify each NYC candidate or employee at least 10 business days before using an AEDT. The notice must disclose:
That an automated tool will be used in the assessment.
The job qualifications and characteristics the tool evaluates.
Instructions for requesting an alternative selection process or accommodation, though the law does not require the employer to grant an alternative.
On request, the data sources and data retention policy for the tool.
The notice can be delivered in the job posting, by email, or through the application process, provided it precedes the tool's use.
What are the penalties for violating Local Law 144?
Penalties are enforced by the DCWP. There is no private right of action under the law itself, so individual candidates cannot sue under Local Law 144 directly, though they may pursue separate claims under federal, state, or city anti-discrimination statutes.
Violation: First violation.
Penalty: Up to $500.
Violation: Each subsequent violation.
Penalty: $500 to $1,500.
Violation: Each day a non-compliant AI tool is used.
Penalty: Each day is treated as a separate violation, allowing penalties to accumulate.
Violation: Failure to provide the required candidate notice.
Penalty: Considered a separate violation in addition to the unlawful use of a non-compliant AI tool.
Because each day and each notice failure can count separately, exposure adds up quickly for high-volume employers. The financial penalty is often secondary to the litigation and reputational risk a documented disparate-impact result can create.
How does Local Law 144 fit with federal and state AI hiring law?
Local Law 144 governs the audit and disclosure mechanics, but it does not displace anti-discrimination law that already applies to algorithmic decisions.
Title VII and the ADA. The EEOC has issued technical guidance confirming that employers remain liable under Title VII for disparate impact caused by algorithmic screening, and under the ADA for tools that screen out applicants with disabilities. A passing Local Law 144 audit does not immunize an employer from a federal disparate-impact claim.
Illinois Artificial Intelligence Video Interview Act. Effective January 1, 2020, it requires consent and disclosure when AI analyzes video interviews, a narrower scope than Local Law 144.
Colorado AI Act (SB 24-205). Creates duties around high-risk AI systems, including those used in employment, centered on preventing algorithmic discrimination and notifying consumers.
EU AI Act. Classifies AI used in recruitment and employment decisions as high-risk under Annex III, triggering conformity, documentation, and human-oversight obligations for systems placed on the EU market.
For a jurisdiction-by-jurisdiction breakdown of these overlapping rules, see our guide to AI hiring laws by state.
What enforcement cases show the stakes?
Real cases establish that algorithmic hiring carries documented legal risk, independent of Local Law 144's audit requirement.
Amazon's internal recruiting tool, reported as scrapped around 2018, down-ranked resumes carrying signals associated with women after training on a male-skewed applicant history. It was never deployed at scale, and it shows how historical data reproduces past bias.
iTutorGroup settled with the EEOC in 2023 for $365,000 after its application-review software automatically rejected female applicants aged 55 and older and male applicants aged 60 and older, an age-discrimination claim under the ADEA.
Mobley v. Workday allowed claims to proceed that an AI screening platform produced age, race, and disability bias, with the court permitting an "agent" theory that a software vendor can be treated as acting for the employers it serves.
These cases sit outside Local Law 144's text but explain why the audit-and-disclosure model exists.
How do you comply with Local Law 144? A step-by-step approach
Use this sequence to move from inventory to documented compliance.
Inventory your tools. List every system that scores, ranks, classifies, or filters candidates for NYC roles, and confirm with each vendor whether the tool meets the AEDT definition.
Map the decision flow. Document whether each tool's output is sole, dominant, or overruling, which determines coverage.
Commission an independent audit. Engage an auditor with no financial stake in the tool to compute selection rates and impact ratios across EEO-1 categories.
Validate the data source. Use historical data where available, and document any reliance on test data.
Publish the summary. Post the audit summary on your public careers page with the distribution date and all required metrics.
Issue candidate notice. Provide the 10-business-day notice describing the tool, the qualifications assessed, and the alternative-process request path.
Set a renewal calendar. Re-audit before the one-year mark and refresh the published summary each cycle.
Keep a federal eye open. Treat a passing audit as a floor, not a defense, and keep validating tools against Title VII and ADA standards.
Frequently asked questions
Does Local Law 144 apply to remote employees?
Yes, when the position is located in New York City or is tied to a NYC office, even if the candidate works remotely. The trigger is the location of the job, not the physical location of the applicant during screening. Employers hiring for NYC-based roles must comply regardless of where applicants sit.
Who is responsible for the bias audit, the employer or the vendor?
The legal obligation rests with the employer or employment agency using the tool, not the vendor. Vendors often commission a shared audit clients can rely on, but the employer remains accountable for confirming it is current, valid, and published before the tool is used in NYC.
Does passing a Local Law 144 audit protect against discrimination lawsuits?
No. A passing audit satisfies the city's audit and disclosure rules but does not bar claims under Title VII, the ADA, the ADEA, or state and city human rights laws. The EEOC has confirmed employers stay liable for disparate impact from algorithmic tools regardless of any city-level audit.
How often must the bias audit be repeated?
At least once every 12 months. The published summary must reflect an audit no older than one year, so employers running AEDTs continuously need an annual re-audit and a refreshed public summary to stay compliant each cycle.
Next steps checklist
Step: 1
Action: Inventory all candidate-scoring and ranking tools used for NYC-based roles.
Owner: HR / Talent.
Step: 2
Action: Confirm AEDT status with each vendor in writing.
Owner: Legal / Procurement.
Step: 3
Action: Engage an independent bias auditor.
Owner: Legal / Compliance.
Step: 4
Action: Verify selection rates and impact ratios across EEO-1 categories.
Owner: Auditor.
Step: 5
Action: Publish the audit summary on the public careers page.
Owner: HR / Web.
Step: 6
Action: Send the 10-business-day candidate notice.
Owner: HR / Talent.
Step: 7
Action: Schedule the annual re-audit and audit summary refresh.
Owner: Compliance.
Step: 8
Action: Cross-check all tools against Title VII and ADA standards.
Owner: Legal.
