Annual bias audit
An independent audit covering selection rates, impact ratios and intersectional cross-tabs across required demographic categories.
A step-by-step compliance guide for NYC Local Law 144. Learn how to inventory AEDTs, conduct bias audits, notify candidates, and prepare for stricter enforcement in 2026.
In December 2025, the New York State Comptroller released a damaging audit of how NYC's Department of Consumer and Worker Protection (DCWP) has been enforcing Local Law 144, the city's regulation on automated employment decision tools (AEDTs).
Independent bias audit, methodology and results review.
Summary on your site, notice to candidates, annual cycle.
An independent audit covering selection rates, impact ratios and intersectional cross-tabs across required demographic categories.
A summary block published on your website covering the audit date, distribution date and source data description.
Each candidate must be notified at least 10 business days before the AEDT is used to evaluate them.
Audits expire after 12 months. New audits before expiration; new audits when models or data materially change.
The findings were stark. Seventy-five percent of test calls to NYC's 311 hotline about AEDT issues never reached DCWP, and when the agency reviewed 32 companies it flagged only one violation โ the Comptroller's own auditors found 17 potential violations in the same set. DCWP has committed to better complaint routing, trained investigators and proactive enforcement, and DLA Piper expects more frequent investigations and higher penalties as a result. If you use AI anywhere in your NYC hiring pipeline, here's what to do now.
An automated employment decision tool is any computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence that issues a simplified output (score, classification, or recommendation) used to substantially assist or replace discretionary decision-making for employment decisions.
In practical terms, this includes:
The key phrase is "substantially assist or replace." If a human decision-maker reviews AI-generated scores or rankings as a significant factor in hiring, that tool is likely an AEDT.
Start by mapping out every tool that touches candidate evaluation. This includes tools you built in-house and tools provided by vendors.
Many companies are surprised to find AEDTs embedded in applicant tracking systems, recruitment marketing platforms, or assessment providers they assumed were purely manual.
What to document for each tool:
If a vendor provides the tool, contact them about their Local Law 144 compliance posture. Note that the law does not impose direct obligations on vendors, so cooperation depends on your contractual relationship.
Local Law 144 applies to employers and employment agencies that use AEDTs to evaluate candidates or employees who reside in New York City. Your company does not need to be based in NYC. If you are hiring for a remote role and a candidate lives in any of the five boroughs, the law applies to that candidate's evaluation.
You are in scope if:
The law requires that an independent auditor conduct the bias audit. DCWP does not maintain an approved list of auditors, so the selection is your responsibility.
What to look for in an auditor:
A vendor can have an independent auditor conduct a bias audit on its tool and share those results with you. If your vendor already has a current audit, you may be able to use it rather than commissioning your own, provided it covers the same tool configuration you deploy.
VerifyWise is one such independent auditor for NYC Local Law 144. We run the bias audit end-to-end and deliver a DCWP-aligned report with the auditor of record listed.
The bias audit must calculate the selection rate (or scoring rate) for each demographic category and the impact ratio, which is the ratio between each group's rate and the group with the highest rate.
Required demographic categories:
How impact ratios work:
If men are selected at a rate of 60% and women at 40%, the impact ratio for women is 40/60 = 0.67 (67%). Under the four-fifths rule, an impact ratio below 80% may signal adverse impact against that group.
The audit must use either historical data from the tool's actual use or test data if historical data is insufficient. The auditor documents findings in a summary report.
VerifyWise runs the bias audit, computes selection rates and impact ratios per the DCWP rules, and delivers a publication-ready summary. One engagement covers the full annual cycle. Get your bias audit.
You must make the bias audit results publicly available on your website. The published summary must include:
There is no prescribed format, but the information must be clearly accessible. Many companies create a dedicated compliance page or add it to their careers section.
Before using an AEDT, you must provide notice to candidates or employees at least 10 business days before the tool is used. The notice must include:
Candidates must also be informed of their right to request an alternative selection process or accommodation, where reasonably available.
Where to provide notice:
The bias audit must be conducted at least once every 12 months. You cannot use an AEDT if more than one year has passed since its most recent audit.
Recommended annual calendar:
| Month | Action |
|---|---|
| Month 1 | Review AEDT inventory for new or changed tools |
| Month 2-3 | Engage auditor and provide data |
| Month 4-5 | Audit conducted and report delivered |
| Month 6 | Publish updated results, review candidate notices |
| Month 7-12 | Monitor for tool changes that require re-audit |
| Ongoing | Log all AEDT usage, maintain candidate notice records |
If you change the AEDT's configuration, retrain a model, or switch vendors mid-cycle, evaluate whether a new audit is needed.
VerifyWise can run it โ full annual cycle in one engagement, 2-4 weeks end-to-end, signed report delivered to your dashboard.
Thinking the law only applies to NYC-based companies. It applies to any employer evaluating candidates who reside in NYC, regardless of where the company is headquartered.
Ignoring vendor-provided tools. If your ATS vendor includes AI-powered screening, that is an AEDT. The compliance obligation falls on you as the employer, not the vendor.
Treating the audit as a one-time event. Audits expire after 12 months. Many companies completed their first audit in 2023 and have not updated since.
Not offering an alternative process. The law requires you to inform candidates they can request an alternative. If you have no alternative available, you should document why and consider establishing one.
Publishing audit results in an inaccessible location. Burying the summary in a PDF behind multiple clicks may not satisfy the public availability requirement. Make it easy to find.
The December 2025 Comptroller's audit matters for two reasons.
First, it showed that DCWP's enforcement has been minimal. Many companies have not faced scrutiny, which may have created a false sense of security.
Second, DCWP has committed to fixing every gap the Comptroller identified. That means improved complaint routing, better-trained investigators, and more proactive enforcement.
The agency will likely move from reactive to actively seeking out non-compliant employers.
Companies that are not yet compliant should treat this as the window to get things sorted before enforcement ramps up.
If you don't have an auditor lined up, VerifyWise can be yours. We run the bias audit end-to-end as the independent auditor of record:
We also map the audit results to related regimes like the EU AI Act for multi-jurisdiction operations.
Curious what a bias audit report looks like? Download a sample NYC Local Law 144 bias audit report generated by VerifyWise, or see the full audit service.
Local Law 144 applies to employers and employment agencies that use automated employment decision tools to evaluate candidates or employees who reside in New York City. Your company does not need to be based in NYC. If you are hiring for a remote role and a candidate lives in any of the five boroughs, the law applies to that candidate's evaluation.
An AEDT is any computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence that issues a simplified output (score, classification, or recommendation) used to substantially assist or replace discretionary decision-making for employment decisions. This includes resume screening tools, video interview analysis platforms, chatbot assessments, candidate matching algorithms, and internal promotion scoring systems.
At least once every 12 months. You cannot use an AEDT if more than one year has passed since its most recent bias audit. If you change the AEDT's configuration, retrain a model, or switch vendors mid-cycle, you should evaluate whether a new audit is needed before continuing to use the tool.
The four-fifths rule comes from EEOC adverse impact analysis. If any demographic group's selection rate is less than 80% of the highest-selected group's rate, that may signal adverse impact against that group. LL144 requires the bias audit to calculate and publish these impact ratios across race, ethnicity, and sex categories. Falling below 0.80 doesn't automatically mean a violation, but it's the threshold that triggers regulatory attention.
The bias audit must be conducted by an independent auditor who has not been involved in using, developing, or distributing the AEDT and who has no employment or financial relationship with the employer that compromises independence. DCWP does not maintain an approved list, so the selection is the employer's responsibility. VerifyWise is one such independent bias auditor for NYC Local Law 144.
Penalties range from $500 to $1,500 per violation, and each day a violation continues is treated as a separate violation. Common violations include failing to conduct a bias audit, failing to provide candidate notice, and failing to publish audit results on the employer's website. The NYC Department of Consumer and Worker Protection enforces compliance.
Yes, if the candidate resides in NYC. The law applies based on where the candidate or employee is located, not where the employer is headquartered or where the role is performed. A remote-first company hiring for a US-based role will be in scope for any candidate who lives in any of the five boroughs.
Yes. VerifyWise runs the bias audit end-to-end as your independent auditor of record โ AEDT inventory, statistical testing for disparate impact across protected categories and intersectional combinations, methodology and results review, and a DCWP-aligned audit report you publish on your site. One engagement covers the full annual cycle.
VerifyWise builds source-available AI governance software used by organizations to manage risk, compliance, and oversight across their AI portfolios. Our editorial team draws on hands-on experience implementing governance workflows for regulated industries and fast-scaling AI teams.
Learn more about VerifyWise โStart your AI governance journey with VerifyWise today.