NYC Local Law 144 (Automated Employment Decision Tools)
Summary
Section titled “Summary”New York City Local Law No. 144 (Local Law 144 of 2021) has been enforced by DCWP since 2023-07-05. It is the first US law with concrete compliance requirements for AI hiring tools. Three core requirements:
- Independent bias audit — a bias audit must be completed within 1 year before using an Automated Employment Decision Tool (AEDT) within New York City.
- Public disclosure — a summary of the audit results must be posted on the company’s website.
- Candidate notice — candidates must be notified at least 10 business days before an AEDT is used, including the tool’s use and the data categories involved.
Definition of AEDT (§ 20-870)
Section titled “Definition of AEDT (§ 20-870)”An “Automated Employment Decision Tool” is:
- derived from machine learning, statistical modelling, data analytics, or artificial intelligence,
- a computational process,
- that issues a simplified output (a score, classification, or recommendation),
- used to substantially assist or replace discretionary decision-making in
- employment decisions (hiring, promotion).
DCWP’s final rule further tightened what counts as “substantial” — the tool must be the sole significant factor in the decision, or replace human decision-making, or override other criteria to be triggered. This tightening drew criticism when the rule took effect in 2023.
Core obligations
Section titled “Core obligations”Independent bias audit
Section titled “Independent bias audit”- Independent auditor: no employment or financial relationship with the employer.
- Statistical metrics: impact ratio, selection rate (broken down by race / ethnicity × sex crosstab).
- Frequency: annual.
- Public disclosure: date of the audit, summary of results, visible on the company’s website.
Candidate notice
Section titled “Candidate notice”- At least 10 business days’ advance notice.
- Content: the tool’s purpose, the categories of data assessed, and data retention.
- Alternative process: candidates may request an alternative assessment process or reasonable accommodation.
Penalties
Section titled “Penalties”- First violation: $500.
- Subsequent: $500-$1,500 per violation, with each day deemed a separate violation.
- The cumulative amount can be substantial (especially for large employers).
Limits and criticisms
Section titled “Limits and criticisms”- Narrowing of “substantial assistance”: the final rule shrank the trigger scope.
- Blind spots in intersectional analysis: intersectional subgroups (e.g., “minority women”) may be too small a sample for statistical significance.
- No private right of action.
- Gap on technology-supplier liability: the obligated entities are employers / employment agencies, not AEDT providers.
Despite these limits, LL 144’s bias-audit + public-disclosure paradigm has been drawn on by the Colorado AI Act and by legislation in Illinois, California, and elsewhere.
Comparison with the Colorado AI Act
Section titled “Comparison with the Colorado AI Act”| Dimension | NYC LL 144 | Colorado AI Act |
|---|---|---|
| Scope | Employment only | 8 categories of consequential decisions |
| Obligated entity | Employer / employment agency | Developer + Deployer |
| Audit | Independent third party | Internal impact assessment; safe-harbour via NIST RMF |
| Candidate / consumer rights | Notice + alternative process | Notice + appeal + human review |
| Penalties | $500–$1,500 / day | Per CPA, no explicit cap |
Primary text and archives
Section titled “Primary text and archives”| Source | Link |
|---|---|
| NYC DCWP page | nyc.gov/site/dca/…/automated-employment-decision-tools |
| Final rule (RCNY Title 6 Chapter 5 Subchapter T) | NYC Rules |
| Statutory text (NYC Admin Code § 20-870 et seq.) | Council |
Revision history
Section titled “Revision history”| Date | Event |
|---|---|
| 2021-12-11 | Mayor signed |
| 2023-04-06 | Final rule issued |
| 2023-07-05 | Enforcement began |