Local Law 144 Compliance Guide: What AI Hiring Tools Need in 2026

By Brendten Eickstaedt — April 14, 2026

Local Law 144 compliance in 2026: who’s in scope, what a bias audit must include, posting + notice rules, and how to avoid common compliance traps.

In Brief

Who must comply: Any employer or agency using AI to screen, score, rank, or recommend NYC candidates, regardless of HQ location. Liability sits on the employer, not the vendor.
Three core obligations: An independent bias audit within 12 months, a public posting of the audit summary on the careers page before use, and written notice to NYC candidates 10+ business days before the tool is used.
What counts as an AEDT: Any tool producing a simplified output (score, tag, classification, ranking, recommendation) used solely, weighted more than any single other criterion, or used to overrule other inputs.
Two scope mistakes: Assuming "the vendor handles it" (they do not own your liability), and over-indexing on company HQ instead of where the role and candidate actually sit. Notice obligations attach to NYC residents.
Selection rate vs scoring rate: yes/no AEDTs use selection-rate analysis, while numerical-score tools with thresholds use scoring-rate analysis. The audit must match the actual decision path.
The impact ratio bar: Selection rate or scoring rate divided by the most-selected/highest-scoring category. The law does not adopt the four-fifths rule, but treat any ratio below 0.80 as a red flag requiring root-cause review.
Auditor independence is defined: Disqualified if they helped develop or sell the tool, have an employment relationship with employer or vendor during the audit, or have direct or material indirect financial interest.
Public posting requirements: Audit date, summary of results with rates and impact ratios for all categories, unknown counts, source/explanation of data used, and the distribution date, your first date of use, not the vendor's release.
Notice as a system-of-record problem: You must prove timing. If you cannot show when notice was delivered relative to AEDT use (by candidate, by channel, with a durable log) you are effectively non-compliant.
Penalties: $375, $1,500 per violation, with each non-compliant candidate counted separately. Violations stack per day.
2026 enforcement shift: NYC Comptroller's December 2025 audit found DCWP missed most non-compliant employers and pushed DCWP toward proactive structured enforcement. Complaint-driven enforcement is over.

Local Law 144 requires every NYC employer using AI in hiring to complete an independent bias audit annually, post a public summary of results, and give candidates at least 10 business days' written notice before the tool is used. In 2026, enforcement is shifting from complaint-driven to proactive after a Comptroller audit exposed widespread gaps in DCWP's compliance checks. In Brief: - Who must comply: Any employer or employment agency using AI to screen, score, rank, or recommend candidates for NYC roles, regardless of where the company is headquartered. Compliance exposure sits on the employer, not the vendor. - The three core obligations: A bias audit by an independent auditor within the last 12 months, a public posting of the audit summary on the careers page before use, and written notice to every NYC candidate or employee at least 10 business days before the tool is used. - What counts as an AEDT: Any tool producing a simplified output (score, tag, classification, ranking, recommendation) that is relied on solely, weighted more than any single other criterion, or used to overrule other inputs. Pure transcription or translation tools are carved out. - Two scope mistakes I see most: Assuming "the vendor handles it" (they do not own your liability), and over-indexing on company headquarters instead of where the role and the candidate actually sit. Notice obligations attach to NYC residents, most ATS systems do not capture residency cleanly. - Selection rate vs scoring rate: If your AEDT makes yes/no advancement decisions, you are in selection-rate territory. If it produces a numerical score with a threshold, you are in scoring-rate territory. The audit must match the actual decision path, most failed audits fail here. - The impact ratio bar: Either selection rate divided by the most-selected category, or scoring rate divided by the highest-scoring category. The law does not adopt the four-fifths rule, but treat any impact ratio below 0.80 as a red flag requiring root-cause review. - Auditor independence is defined: An auditor is disqualified if they helped develop or sell the tool, have an employment relationship with employer or vendor

This is a free preview. Upgrade to Pro to read the full article.