NYC Local Law 144: The Complete 2026 AEDT Bias Audit Guide

NYC Local Law 144 has been the quiet-but-real US frontline rule for AI in hiring since DCWP began enforcement on July 5, 2023. For almost three years, enforcement was light. That changed at the end of 2025 when the New York State Comptroller published an audit calling out DCWP for under-enforcement, and again in early 2026 when DCWP announced a new enforcement-forward posture under Commissioner Sam Levine. This guide covers what an AEDT actually is, what counts as a compliant bias audit, what candidates and the public need to see, and what penalties look like now that DCWP is leaning in.

If you screen job candidates in New York City with anything more sophisticated than a keyword filter, this is your law.

When the law actually started biting

DCWP began enforcement of Local Law 144 of 2021 on July 5, 2023 (NYC DCWP). The final rules adopted on April 6, 2023 broadened the scope significantly from the proposed version, particularly around what counts as an automated employment decision tool. That April rulemaking removed the requirement that inputs and parameters be refined through cross-validation or training/testing data, which pulled statistical models and even some simpler scoring tools into the AEDT definition (Hunton Andrews Kurth).

What counts as an AEDT

An Automated Employment Decision Tool is a computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence that issues simplified output, including a score, classification, or recommendation, used to substantially assist or replace discretionary decision-making in employment decisions including hiring or promotion (Hunton Andrews Kurth).

Two phrases do most of the work: "simplified output" and "substantially assist or replace discretionary decision-making." If your tool produces a rank, score, classification, or shortlist that recruiters routinely act on without independently re-evaluating each candidate, you are in. The rule explicitly covers resume-screening tools and interview-scheduling tools used even at the earliest hiring stages, well before a final decision (Hunton Andrews Kurth).

What does not count: tools where a human reviewer independently re-evaluates each candidate without weighting the AEDT output disproportionately. The rule treats "substantial assistance" as the question. If your hiring manager has the model output on screen and uses it as the primary signal, you are running an AEDT.

The bias audit: who can do it

The bias audit must be conducted by an independent auditor. Independent means three things: (1) the auditor was not involved in developing the AEDT, (2) the auditor is not involved in using the AEDT for employment decisions, and (3) the auditor has expertise to conduct bias audits (nycbiasaudit.com).

The "not involved in development" prong rules out the vendor's own employees. The "not involved in use" prong rules out the employer's own HR or analytics team. That leaves third-party auditors, of which there are now several specialising in LL 144 work, including Warden AI, BABL AI, Holistic AI, and Eightfold's referenced audit partners.

An employer may rely on a bias audit conducted using another employer's historical data only if the employer either provided its own historical data to the auditor or has never used the AEDT (Hunton Andrews Kurth). That carve-out matters. It allows a vendor to commission one audit per AEDT and have employer-clients ride on it, as long as the employer either contributed data or is a brand-new user.

What the bias audit measures

The audit calculates selection rates for hiring/promotion tools and scoring rates for tools that output scores, broken out by race, sex, and ethnicity categories (Hunton Andrews Kurth).

The headline metric is the impact ratio: each group's selection or scoring rate divided by the rate for the group with the highest rate. An impact ratio below 0.80, the four-fifths rule from EEOC's Uniform Guidelines on Employee Selection Procedures, generally indicates potential adverse impact (nycbiasaudit.com / Eightfold sample audit).

Published results must include: selection or scoring rates for all categories, the number of individuals the AEDT assessed in each category, the number of individuals not included because they fall into an "unknown" category, and the audit date (Hunton Andrews Kurth).

Four notes on the audit that trip people up:

• The bias audit must be conducted annually, within one year prior to each use of the AEDT (NYC DCWP). An old audit does not cover new hiring cycles.
• Sub-group analysis (intersectional, like Black women) is encouraged but not strictly required when sample sizes are small.
• The "unknown" demographic bucket is a real category. The summary must disclose how many candidates fell into it.
• Historical data audits require at least one year of actual data. If the employer has none, the AEDT can be audited using the vendor's pooled employer data, subject to the carve-out above.

Candidate notice: 10 business days

Employers must provide notice to candidates at least 10 business days prior to use of an AEDT (NYC DCWP). The notice must state (i) that an AEDT will be used, (ii) how it will be used, and (iii) the data that will be collected.

Candidates must be offered an alternative selection process or reasonable accommodation upon request (nycbiasaudit.com). In practice, most large employers handle this with a checkbox in the job application acknowledging that an AEDT may be used, paired with a link to a more detailed policy page. The simpler approach is to post the notice in the job listing itself.

The 10-business-day window is not flexible. If you launch a job posting on a Monday and someone applies that day, you cannot run them through an AEDT before the following Monday at the earliest. That has a meaningful operational impact on high-volume hiring funnels.

Public summary of audit results

A summary of the bias audit must be publicly available on the employer's website or linked from the job posting. Required contents: selection rates and impact ratios for each protected category, sample size for each category (including the "unknown" counts), and the audit date (Hunton Andrews Kurth).

The summary is what plaintiffs, journalists, and regulators read first. It is also what other prospective candidates see. If your impact ratio is below 0.80 for any category, expect questions. The audit framework does not penalise you for publishing a sub-0.80 ratio. It penalises you for not auditing and not disclosing.

Penalties

Civil penalties are $500 per violation for a first offence and $1,500 per violation for each subsequent violation, per day of non-compliance (NY State Comptroller / nycbiasaudit.com).

Per-day stacking is the part to watch. A single job listing that violates the notice rule for thirty days is potentially thirty separate violations. A continuous AEDT deployment that fails the audit-currency requirement can accrue quickly. There is no statutory cap; aggregate exposure is the product of duration and the number of distinct AEDT uses.

The State Comptroller audit that changed the temperature

On December 2, 2025, the New York State Comptroller published an audit of DCWP's enforcement of Local Law 144 covering July 2023 through June 2025 (NY State Comptroller). The findings were unflattering.

DCWP had received only 2 AEDT complaints in the audit window. The Comptroller's own review of 32 companies found just one issue of non-compliance documented by DCWP, even though the Comptroller's auditors identified at least 17 potential violations across the same sample. The complaint intake process was assessed as ineffective. The Comptroller issued 13 recommendations. DCWP agreed to adopt 10 of them, including stronger enforcement, cross-training of staff on AEDT technical issues, and consulting with the Office of Technology and Innovation for technical expertise (DCI Consulting).

DCWP's enforcement-forward posture in 2026

Under new DCWP Commissioner Sam Levine, formerly the director of the FTC's Bureau of Consumer Protection, DCWP shifted to an enforcement-forward posture in early 2026 (Gibson Dunn). That is a meaningful change. Levine ran the FTC bureau that built the deception-and-unfairness case file against major tech platforms. He arrives at DCWP with a clear mandate to use the agency's investigative and rulemaking authority more aggressively.

DCWP's FY2026 Regulatory Agenda, published in May 2025, did not list amendments to the LL 144 rules themselves, but the agency publicly committed to strengthening enforcement practices (NYC Rules FY2026 Agenda PDF). Translation: same rules, more enforcement.

For employers, the practical implication is that the historical safety of an unaudited or under-audited AEDT has expired. Expect DCWP to begin proactive sweeps of job postings in 2026 and to bring at least one high-profile enforcement action to set the tone.

What an LL 144 compliance programme should actually contain

For employers using AEDTs in NYC, the operating checklist is short but disciplined.

1. Inventory every tool that touches NYC hiring or promotion decisions. Be expansive. Include third-party platforms, sourcing tools, and assessment vendors.
2. For each tool, confirm whether it meets the AEDT definition. If yes, identify the most recent bias audit. Confirm it is within the last 12 months.
3. If the audit relied on the vendor's pooled employer data, confirm your eligibility under the carve-out. Either you have not used the tool before, or you supplied your own historical data.
4. Publish the audit summary on a stable URL. Link to it from every NYC job posting that uses the tool, or from a central careers compliance page.
5. Add the 10-business-day candidate notice to every relevant job listing. Include the language about how the AEDT is used and what data is collected. Offer the alternative-process route.
6. Set a calendar reminder for the next annual audit. Treat the renewal as a board-level compliance milestone.

For vendors selling AEDTs into NYC employers, the bar is to commission a credible, independent annual audit, publish it on a stable URL, and provide your customers with a one-page compliance pack containing the audit summary, the candidate notice template, and the alternative-process language.

How LL 144 fits the broader US picture

LL 144 is the most-cited and most-litigated of the US algorithmic hiring laws, but it is not alone. Illinois's AI Video Interview Act, Maryland's HB 1202, and California's emerging AB 2930 framework all overlap with the LL 144 territory. The EU AI Act treats employment-AI as Annex III high-risk under the Digital Omnibus extension, with the December 2, 2027 deadline. For employers operating cross-jurisdictionally, an LL 144-compliant programme is the right floor, not the ceiling. See our employment HR framework page and our NYC Local Law 144 framework summary for the broader landscape.

References

1. NYC DCWP. Automated Employment Decision Tools. https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page
2. Hunton Andrews Kurth. NYC DCWP Adopts Rules to Implement Law Governing Automated Employment Decision Tools. April 26, 2023. https://www.hunton.com/privacy-and-cybersecurity-law-blog/nyc-dcwp-adopts-rules-to-implement-law-governing-automated-employment-decision-tools-and-sets-july-enforcement-date
3. NY State Comptroller. Enforcement of Local Law 144 — Automated Employment Decision Tools. December 2, 2025. https://www.osc.ny.gov/state-agencies/audits/2025/12/02/enforcement-local-law-144-automated-employment-decision-tools
4. DCI Consulting. Recommended Changes to Local Law 144. February 24, 2026. https://blog.dciconsult.com/recommended-changes-to-local-law-144
5. nycbiasaudit.com. How to Comply with the NYC Bias Audit Law. https://www.nycbiasaudit.com/blog/how-to-comply-with-the-nyc-bias-audit-law
6. Gibson Dunn. New York City DCWP Adopts Aggressive Enforcement Posture Under New Leadership. April 8, 2026. https://www.gibsondunn.com/new-york-city-dcwp-adopts-aggressive-enforcement-posture-under-new-leadership/
7. Eightfold AI bias audit summary. https://eightfold.ai/wp-content/uploads/eightfold-summary-of-bias-audit-results.pdf
8. Warden AI. NYC LL 144 explainer. https://www.warden-ai.com/resources/nyc-ll-144
9. NYC Rules. DCWP FY 2026 Regulatory Agenda. https://rules.cityofnewyork.us/wp-content/uploads/2025/05/DCWP-FY-2026-Regulatory-Agenda-Final.pdf