NYC Local Law 144 AI Bias Audit: What HR Tech Companies Must Do in 2026
New York City is enforcing the first municipal law in the US requiring mandatory third-party bias audits for AI hiring tools. If your product screens NYC candidates, compliance is not optional — and each day of non-compliance is a separate fine.
What Is NYC Local Law 144?
NYC Local Law 144 (formally Int. 1894-A) went into effect on January 1, 2023, with enforcement beginning July 5, 2023. It is the first US law to require mandatory independent bias audits for AI tools used in employment decisions.
The law targets what it calls Automated Employment Decision Tools (AEDTs) — any computational process that substantially assists or replaces human decision-making in employment screening. If your HR tech product uses AI to rank, score, or filter candidates, it almost certainly qualifies as an AEDT.
The law applies to both vendors who sell AEDTs and employers who use them to screen candidates for positions located in New York City. A Chicago-based HR tech startup selling a resume-screening AI to NYC employers is in scope for the vendor obligations.
What Counts as an AEDT?
The NYC DCWP (Department of Consumer and Worker Protection) has issued guidance on what qualifies as an AEDT. The definition is intentionally broad:
- Resume screening algorithms — AI that scans resumes and decides which candidates advance (or assigns a score that influences that decision)
- Video interview analysis tools — AI that analyzes facial expressions, tone of voice, word choice, or other video/audio signals from recorded interviews to score candidates
- Cognitive and personality assessments — AI-scored pre-employment tests that produce a pass/fail or ranked score
- Skills-matching platforms — AI that matches candidates to open roles based on inferred skills or profile similarity
- Chatbot pre-screeners — AI chatbots that conduct screening interviews and make pass/fail recommendations
Importantly, the law requires that the tool "substantially assist or replace discretionary decision-making." Tools that merely surface information to a human without making a judgment — like an ATS that displays applications without ranking them — may fall outside scope. But if your tool produces a score, ranking, or recommendation that a recruiter is likely to follow, it's almost certainly an AEDT.
The Three Compliance Requirements
Independent Bias Audit
REQUIRED ANNUALLYAn independent third-party auditor must conduct a bias audit of the AEDT within one year before the tool is used. The audit must calculate the selection rate and impact ratio for sex, race/ethnicity, and intersectional categories (e.g., Black women). The auditor must be independent — meaning not employed by the vendor or employer.
The audit results must be published publicly on the employer's or vendor's website, with the date of the most recent audit, the source of data used, and the number of candidates assessed. Results must remain posted for at least 2 years.
Candidate Notice
BEFORE USEEmployers must notify candidates in NYC that an AEDT will be used to evaluate them. Notice must be given at least 10 business days before the tool is used, and must describe the AEDT and the job qualifications/characteristics it evaluates.
Candidates must also be informed of their right to request an alternative selection process — the employer must provide a reasonable accommodation if the candidate objects to the AEDT screening.
Data Retention
2-YEAR MINIMUMEmployers must retain records of their AEDT use, the bias audit results, and candidate notification records for a minimum of two years. NYC DCWP can request these records during an investigation.
For vendors: bias audit summaries must be publicly accessible on your website for 2 years from the date the audit is no longer being relied upon.
What a Local Law 144 Bias Audit Actually Measures
The audit methodology is prescribed in the law. Auditors must calculate:
Selection Rate
The proportion of applicants in a given sex/race category who are selected (or advanced) by the AEDT. Calculated per category.
Impact Ratio
The selection rate for a category divided by the selection rate for the category with the highest selection rate. A ratio below 0.80 (the '4/5ths rule') typically indicates adverse impact against that group.
Categories Required
Sex (male/female), Race/Ethnicity (EEOC categories: Hispanic/Latino, White, Black/African American, Native Hawaiian/Pacific Islander, Asian, American Indian/Alaska Native, Two or more races), and intersectional combinations (e.g., Hispanic women, Black men).
Unknown Category Handling
Candidates who decline to provide demographic info are assigned to 'Unknown' category. The audit must include the 'Unknown' group in calculations. If more than 2% of the scored population is Unknown, the auditor must note this limitation.
The audit does not require a specific pass/fail threshold — a tool is not automatically "banned" if it shows disparate impact. But the results must be published, and employers can face discrimination claims under existing civil rights law if the tool shows significant adverse impact and the employer continues using it without justification.
Penalties for Non-Compliance
NYC DCWP enforces Local Law 144. Violations carry civil penalties of:
$375 per day
Minimum civil penalty for each day of non-compliance
$1,500 per day
Maximum civil penalty per violation per day
Per violation
Each unlawful use of an AEDT can be a separate violation
Private lawsuits
Candidates have a private right of action under NYC Human Rights Law for discriminatory use
These fines accumulate quickly. An employer using a non-compliant AEDT for 90 days faces potential liability of $135,000+ — and that's before any civil rights claims from affected candidates.
Who Needs to Act: Vendors vs. Employers
HR Tech Vendors
- ▸Get an independent bias audit of your tool using a representative dataset
- ▸Publish audit summary on your website — including selection rates, impact ratios, and audit date
- ▸Re-audit annually (or more frequently if you make significant changes to the model)
- ▸Contractually require employers who use your tool in NYC to comply with their obligations
- ▸Provide employers with the demographic data collection tools needed for candidate notice
Employers Using AEDTs
- ▸Confirm your vendor has a current bias audit — ask for the public posting link before deployment
- ▸Notify NYC candidates at least 10 business days before AEDT screening
- ▸Provide an alternative selection process for candidates who opt out of AEDT screening
- ▸Retain all AEDT use records and candidate notifications for 2 years
- ▸Update job postings to disclose use of AEDTs in the screening process
Local Law 144 Compliance Checklist
For HR tech vendors — complete before deploying any AI screening tool for NYC-based hiring.
The Bigger Picture: Algorithmic Hiring Laws Are Spreading
NYC Local Law 144 is the leading edge of a broader movement. Multiple US jurisdictions have passed or are considering algorithmic hiring regulations:
AI Video Interview Act (2020)
Requires employers to notify candidates before using AI to evaluate video interviews, explain how the AI works, and obtain consent. Results cannot be shared without candidate approval.
HB 1202 (2020)
Requires employers using facial recognition in job interviews to obtain written consent before use.
CCPA + AB 331 (pending)
CCPA gives candidates rights over automated processing of their personal data. AB 331 would require algorithmic discrimination audits for high-risk automated decision systems.
SB 205 (2024)
Governs high-risk AI systems including employment decisions. Requires developers and deployers to disclose AI use and provide notice to affected consumers, with anti-discrimination protections.
EU AI Act — High Risk
AI used in employment and HR management is classified as high-risk under the EU AI Act, requiring full conformity assessments, human oversight, and registration. Enforceable August 2026.
The compliance burden for HR tech companies is growing. Local Law 144 is the most prescriptive US law today, but it won't be the last. Building bias audit infrastructure now — independent auditing relationships, demographic data collection flows, public result posting — positions you ahead of the expanding regulatory curve.
Monitor your HR tech product for compliance exposure
RatedWithAI helps tech teams track compliance requirements across their web properties and products. Start with a free scan to understand your current compliance posture.
Scan Your Product for Free →Frequently Asked Questions
Does Local Law 144 apply to companies outside New York City?
Yes, if your AI tool is used to screen candidates for positions located in NYC. The law applies based on where the job is located, not where the employer or vendor is headquartered. A San Francisco HR tech startup whose tool is used by a NYC employer to screen NYC-based applicants is in scope for vendor compliance obligations.
What if I use a third-party AI hiring tool — am I responsible too?
Both the employer and the vendor share responsibilities. As the employer, you must: confirm the vendor has a current public bias audit, notify candidates before using the tool, and provide an alternative selection process on request. You cannot shift all liability to the vendor — if the tool has disparate impact and you continue using it, you face exposure under NYC Human Rights Law.
How often does the bias audit need to be updated?
The bias audit must be no more than one year old at the time the AEDT is used. If you make significant changes to the AI model — retrain it, change the features, adjust scoring thresholds — you should re-audit before re-deploying. Annual re-auditing is the practical minimum even without model changes, because the user population and underlying data may shift.
What if the bias audit shows disparate impact? Is the tool banned?
No — Local Law 144 does not set a threshold at which a tool is prohibited. The law requires disclosure of bias audit results, including any adverse impact found. However, if results show significant disparate impact and the employer continues using the tool, they may face discrimination claims under existing federal and NYC civil rights law. The audit is both a compliance requirement and a legal discovery document.
Is a resume keyword filter an AEDT?
It depends on how it works. A simple keyword search that returns results for human review is unlikely to qualify. A system that scores or ranks resumes based on keywords, automatically removes candidates who don't hit a threshold, or uses ML to predict candidate quality — that almost certainly qualifies as an AEDT. When in doubt, treat it as an AEDT.