# NYC AI Hiring Law: Compliance Requirements for AI Recruiting Tools
New York City's Local Law 144 (effective April 2023) created the first major US regulation specifically targeting AI in hiring. This guide explains what the law requires and how to comply.
Law Overview
What LL144 Covers
Automated Employment Decision Tools (AEDTs) The law applies to any computational process that: - Uses machine learning, statistical modeling, data analytics, or AI - Issues simplified outputs (scores, classifications, recommendations) - Substantially assists or replaces discretionary decision-making
Covered Decisions - Screening candidates for employment - Screening employees for promotion
Who Must Comply
Employers in NYC who use AEDTs for: - Hiring decisions (any role based in NYC) - Promotion decisions (employees based in NYC)
Employment Agencies operating in NYC
Note: The law applies based on job location, not employer headquarters. A company in California hiring for a NYC role must comply.
> Download our free AI Recruitment Playbook — a practical resource built from real implementation experience. Get it here.
## Core Requirements
Requirement 1: Annual Bias Audit
Audit Scope - Must cover all AEDTs used in employment decisions - Must be conducted by independent auditor - Must be completed within one year prior to use - Must be updated annually
Audit Content
The audit must calculate and report:
Selection Rate Analysis For each AEDT, calculate selection rate (or scoring rate) by: - Sex categories (male, female) - Race/ethnicity categories (Hispanic/Latino, White, Black/African American, Native Hawaiian/Pacific Islander, Asian, Native American/Alaska Native, Two or more races) - Intersectional categories (sex × race/ethnicity)
Impact Ratio Calculation ``` Impact Ratio = Selection Rate of Group / Selection Rate of Most Selected Group ```
Example: - Female selection rate: 40% - Male selection rate: 50% - Impact ratio (female): 40%/50% = 0.80
Audit Report Requirements - Date of audit - Source and explanation of data used - Number of individuals assessed - Selection/scoring rates by category - Impact ratios by category
Requirement 2: Public Posting of Audit Summary
What to Post - Summary of most recent bias audit - Distribution date of the AEDT - Must be posted on company website
Where to Post - Clear and conspicuous location - Employment section of website recommended - Must remain posted during AEDT use
Requirement 3: Candidate/Employee Notice
Notice Timing - At least 10 business days before AEDT use - Can be provided at any point in process (application, interview scheduling, etc.)
Notice Content - That an AEDT will be used - Job qualifications/characteristics the AEDT will assess - Information about data retention policy - How to request alternative selection process (if available) - How to request reasonable accommodation
Notice Methods - Email to candidate/employee - Statement in job posting - Prominent notice on careers website
Requirement 4: Data Access Rights
Upon Request, Provide - Category of data collected for AEDT - Source of data (if not from candidate) - Data retention policy
Audit Methodology Details
Independent Auditor Requirements
The law requires an independent auditor but doesn't specify credentials. Best practices: - No financial interest in tool's success - Relevant statistical/technical expertise - Employment law knowledge helpful - I/O psychology background valuable
Data Requirements for Audit
Historical Data (Preferred) - Use actual selection data from past year - Must have sufficient sample sizes - Need demographic data (often from applicant flow logs)
Test Data (Alternative) - If historical data unavailable or insufficient - Use representative sample data - Must reflect likely applicant pool
Intersectional Analysis
The law requires intersectional analysis (sex × race/ethnicity). This creates many categories:
| Category | Example |
|---|---|
| Male × White | White males |
| Male × Black | Black males |
| Female × Asian | Asian females |
| ... | ... |
Sample Size Challenges Some intersectional categories may have small samples. The rules permit: - Excluding categories with fewer than 2% of total data - Noting sample size limitations in report
Recommended Reading
- AI Recruitment: How Companies Are Reducing Time-to-Hire 63% While Improving Quality of Hire
- The Complete AI Hiring Bias Audit Checklist for HR Leaders
- AI Performance Management: Moving Beyond Annual Reviews
## Compliance Implementation
Step 1: AEDT Inventory
Identify all tools that might qualify as AEDTs: - [ ] Applicant tracking system AI features - [ ] Resume screening tools - [ ] Video interview analysis - [ ] Assessment platforms with AI scoring - [ ] Chatbots making screening decisions - [ ] Internal promotion algorithms
Step 2: Vendor Assessment
For each identified AEDT: - [ ] Does vendor provide LL144 compliance documentation? - [ ] Does vendor have independent bias audit? - [ ] What data does vendor collect? - [ ] What is vendor's data retention policy?
Step 3: Data Preparation
Prepare for bias audit: - [ ] Compile demographic data for applicants/employees - [ ] Document selection decisions by AEDT - [ ] Ensure data quality and completeness - [ ] Plan for intersectional analysis
Step 4: Conduct Audit
- [ ] Select independent auditor
- [ ] Provide required data
- [ ] Review draft results
- [ ] Address any issues identified
- [ ] Finalize audit report
Step 5: Notice Implementation
- [ ] Update job postings with AEDT notice
- [ ] Modify application process for notice timing
- [ ] Create data access request process
- [ ] Post audit summary on website
Step 6: Ongoing Compliance
- [ ] Schedule annual audit renewal
- [ ] Monitor regulatory updates
- [ ] Track enforcement trends
- [ ] Update processes as needed
Enforcement and Penalties
Enforcement Agency
NYC Department of Consumer and Worker Protection (DCWP)
Penalties
| Violation | First | Subsequent |
|---|---|---|
| Failure to conduct bias audit | $500 | $500-$1,500 |
| Failure to publish audit summary | $500 | $500-$1,500 |
| Failure to provide notice | $500 | $500-$1,500 |
Each day constitutes a separate violation.
Per-Candidate Risk: Each candidate not properly noticed could trigger separate violation.
Enforcement Approach
DCWP has indicated: - Initial focus on education over enforcement - Complaints will be investigated - Audits of employers may occur - Patterns of non-compliance will be prioritized
Common Compliance Questions
Q: Does screening by keywords without AI count?
A: Likely no. The law requires machine learning, statistical modeling, or AI. Simple keyword matching without scoring or learning is probably not covered. However, if keywords feed into a scoring algorithm, it likely is covered.
Q: Does human review eliminate AEDT classification?
A: Not necessarily. If the tool "substantially assists" the decision, human review doesn't exempt it. If humans truly make independent decisions and the tool is just one factor, there may be an argument for exemption.
Q: What about tools used before the interview?
A: All stages are covered. Whether screening resumes, scheduling interviews, or scoring assessments, if an AEDT is used for any employment decision stage, it's covered.
Q: Do we need separate audits for each AEDT?
A: The audit must cover each AEDT used. A single audit report can cover multiple tools, but each must be analyzed separately.
Beyond NYC: Preparing for Future Regulation
NYC is first, not last. Similar laws are emerging: - California proposed legislation - Illinois amendments - Colorado AI transparency requirements - EU AI Act (high-risk classification for employment AI)
Preparation Strategy - Build compliance infrastructure now - Document all AI hiring tools - Establish bias monitoring practices - Create scalable notice processes
Contact APPIT's HR technology team for LL144 compliance assistance.
