NYC Local Law 144 mandates an independent bias audit before any automated tool screens job candidates or employees for promotion in New York City. The audit must be completed each year and the auditor cannot be your vendor, your consultant, or anyone with a financial stake in the outcome.
New York and California have each taken a different approach to governing AI in employment but both reach the same core issue: whether automated tools produce outcomes that fall unequally across protected groups.
Any employer or employment agency that uses an AEDT to screen candidates for positions in NYC, or employees for promotion, must obtain an independent bias audit before use and repeat it annually. The law applies even when the tool is provided by a vendor the employer bears the duty.
The audit must calculate selection rates and impact ratios by sex, race/ethnicity, and intersectional categories. Demographic data cannot be inferred or imputed. Results are posted publicly on the employer's website for at least six months after the tool's last use.
California's CCPA regulations extend privacy and fairness obligations to ADMT technology that processes personal information and uses computation to replace or substantially replace human decision-making. Employment decisions are explicitly within scope.
Businesses must provide pre-collection notice of ADMT use, conduct risk assessments documenting the system's logic and data types, and give consumers the right to opt out of significant decisions. A qualifying human appeal process can substitute for opt-out rights in hiring and promotion contexts.
The law's definition is broader than many employers expect. It is not limited to sophisticated AI or machine learning a tool qualifies based on how it is used, not what it is built on.
A tool is an AEDT when its output is used as the sole basis for a decision, is weighted more heavily than other criteria, or overrides a conclusion including a recruiter's judgment. The weight matters, not the label.
If your ATS vendor, assessment platform, or resume-screening software qualifies as an AEDT and you use it to screen NYC candidates, the compliance obligation belongs to you not to the software company. Contracts do not transfer the duty.
The law covers existing employees being screened for promotion, not only external candidates. An internal talent tool, a performance ranking system, or a promotion-eligibility filter may trigger the requirement.
The law specifies what the audit must calculate. Our deliverables go further giving you documentation that survives regulatory scrutiny and public posting that candidates can actually understand.
Calculated by sex, race/ethnicity, and intersectional categories. Includes count of individuals with unknown demographic data, documented per regulatory requirement.
A compliant summary formatted for posting on your careers or corporate responsibility page, including audit date, auditor identity, and summarized impact ratios.
Documentation of the data source, collection period, and method required by the law and necessary for subsequent audits to confirm consistency year over year.
A signed statement confirming the auditor meets LL 144's independence requirements no employment relationship, no role in developing the tool, no material financial interest in the employer or vendor.
Where impact ratios fall below thresholds established by existing civil rights guidance, we document the finding and its regulatory significance without inferring discrimination, which the law does not authorize auditors to conclude.
For clients who want to understand what drives observed disparities, we can extend the engagement to examine model inputs, scoring distributions, and candidate pool characteristics this work is separate from the compliance audit itself.
The law is specific: the auditor cannot have an employment relationship with you or your AEDT vendor, cannot have played any role in developing or deploying the tool, and cannot have a material financial interest in either party. Many "audits" offered by AI platform vendors or by consultants with ongoing relationships with those vendors do not clear this bar.
We do not resell, refer, or white-label any AI hiring platform. Our revenue comes from the audit, not from the tool we're auditing. That's what independence means in practice.
John Smutniak and our analytics team have decades of experience in statistical analysis and measurement not in AI ethics generalism. We calculate impact ratios against actual data, not modeled estimates.
As Certified Lead Auditors for ISO 42001, ISO 27001, and ISO 27701, we bring formal audit discipline to LL 144 work structured evidence review, documented methodology, and a report that holds up under regulatory scrutiny.
State and municipal AI hiring laws are multiplying. What satisfies NYC LL 144 today won't cover everything on the horizon. A documented audit program built around LL 144's requirements is a defensible foundation for the rules still taking shape.
| Jurisdiction | Law / Rule | Scope | Status |
|---|---|---|---|
| New York City | Local Law 144 (AEDT) | Annual bias audit, public posting, candidate notice before use | In Force |
| California | CPPA ADMT Regulations | Risk assessments, consumer notice, opt-out rights for significant decisions | Enacted · Jan 2027 |
| Illinois | AI Video Interview Act | Disclosure, consent, and data retention limits for AI-analyzed video interviews | In Force |
| Maryland | HB 1202 (Facial Recognition) | Restricts facial recognition use in employer hiring screening | Enacted |
| EU (importing) | EU AI Act Annex III | High-risk AI systems include recruitment and employment tools; conformity assessment required | In Force Aug 2026 |
A bias audit under LL 144 tells you the selection rates and impact ratios. It does not tell you why they are what they are, or whether your tool is worth keeping. That's a harder question and it's the one most employers haven't started asking.
John Smutniak · Managing Partner, MeasurementPros
We work from the data your tool has already produced. No software integration, no access to your ATS, no new vendor relationship.
We confirm the tool qualifies as an AEDT, identify what historical data exists, and outline the documentation we need. If your data has gaps we can't work around, we tell you before the engagement starts.
You provide the historical selection data typically a file export from your vendor along with available demographic data (self-identified, not inferred) and documentation of how the tool is used in your screening process.
We calculate selection rates and impact ratios by sex, race/ethnicity, and intersectional categories. If the data requires clarification, we request it we do not fill gaps with assumptions.
We deliver the full audit report and a public summary formatted for posting. The public summary covers what the law requires audit date, auditor, tool description, and impact ratios. The full report is yours; what you post is governed by LL 144.
LL 144 requires the audit be repeated each calendar year. We track your renewal date, notify you 90 days in advance, and carry your data methodology forward to ensure year-over-year comparability.
Photo coming
Who does this work
Managing Partner · Certified Lead Auditor ISO 42001, ISO 27001, ISO 27701
John has worked in measurement, risk analytics, and controls for over two decades. The statistical methodology behind our LL 144 audits is his core discipline impact ratio analysis rooted in psychometrics and quantitative research, not repurposed software compliance checklists. Independence from vendors is structural, not a policy we adopted.
Tell us about your tool and timeline. We'll confirm scope and send a data checklist within two business days.