The NYC bias audit law (Local Law 144):
What it requires and how to comply
If you use software to help screen, score, or rank job candidates, and any of those candidates live in New York City, there is a good chance a New York City law already applies to you. It’s called Local Law 144, usually shortened to the AEDT (Automated Employment Decision Tools) law or the bias audit law, and the city has enforced it since July 2023.
The law requires three things: that you have your hiring tool checked for bias by an independent auditor, that you publish a summary of what the audit found, and that you tell candidates the tool is being used before it’s used on them. The penalties are per-violation and accumulate: each day you use a covered tool without a valid audit, and each candidate you fail to notify counts as a separate violation.
What is NYC Local Law 144, the bias audit law?
Local Law 144 is New York City's rule for automated employment decision tools. It is enforced by the city's Department of Consumer and Worker Protection (DCWP). It took effect on January 1, 2023, and the city began enforcing it on July 5, 2023.
The law applies to employers and employment agencies that use these tools to make hiring or promotion decisions about people in New York City. It doesn’t regulate the companies that build the tools; the obligation falls on the employer or agency using it.
It also reaches well beyond New York-based companies. Because the law follows the candidate rather than the company, an employer anywhere in the world is covered if it uses a qualifying tool to evaluate someone who lives in the city. A fully remote role open to NYC residents can be enough. The law is limited to hiring and promotion, and does not apply to decisions about pay, discipline, or termination.
What counts as an automated employment decision tool?
The law defines an automated employment decision tool, or AEDT, as any computational process, built from machine learning, statistical modeling, data analytics, or artificial intelligence, that produces a simplified output such as a score, a rating, a ranking, or a recommendation, and that is used to substantially assist or replace a human decision in hiring or promotion.
In practical terms, the tools most likely to be covered include:
-
Resume screeners that filter or rank applicants,
-
Assessment tools that score candidates for fit or predicted performance,
-
Systems that grade recorded video interviews,
-
Chat- or game-based assessments that produce a score used in selection.
Tools that only transcribe or translate text are specifically excluded, for example software that converts a resume to plain text, or transcribes an interview for a person to read, is not an AEDT on its own.
The phrase "substantially assist or replace" has a specific meaning, and it determines whether you are covered at all. Your use of a tool qualifies if the tool's output is the only factor in a decision, if it carries more weight than any other single factor, or if it can overrule a conclusion that a person reached. If a recruiter weighs the output alongside other factors and it is not the deciding one, the use may sit outside the law. This is a judgment worth making carefully and documenting, because it determines whether you have to do anything at all.
What the law requires
Local Law 144 has three requirements, and all three must be in place before you use a covered tool:
-
An independent bias audit of the tool, no more than one year old,
-
A public summary of the audit's results, posted where candidates can see it,
-
Advance notice to candidates and employees that the tool is being used.
How the bias audit works
At its core, the bias audit is a test for disparate impact: whether the tool selects or favors some groups at meaningfully different rates than others. An independent auditor runs the analysis on two characteristics, for example sex and race or ethnicity, using the categories from the federal EEO-1 report. It also covers the intersection of the two, such as Hispanic women or Black men, which produces fourteen combined groups.
For each group, the auditor calculates an impact ratio. For tools that select or reject candidates, the impact ratio is the group's selection rate divided by the selection rate of the most-selected group. For tools that produce a continuous score, the auditor first finds the median score across everyone assessed, then compares the share of each group scoring above it. The benchmark behind all of this is the long-standing four-fifths rule from federal employment law, which treats a ratio below 0.80 as a sign of possible adverse impact.
One widely misunderstood detail is that the law requires you to calculate and publish these ratios. It does not require your tool to reach any particular number. There is no passing score, and the law doesn’t force you to change a tool that shows disparities, only to disclose them. Publishing poor numbers can carry its own risk under other anti-discrimination laws, which is a good reason to understand your results before they are public, but weak numbers alone are not a violation of Local Law 144.
The audit normally runs on your historical data, meaning the real results from your own use of the tool. If you do not have enough data to produce a statistically meaningful result, for example because the tool is new to you, the auditor may use test data instead, as long as the published summary explains why and how that data was produced. An auditor may set aside any group that makes up less than 2% of the data, but must still report that group's numbers.
Who counts as an independent auditor
The auditor has to be genuinely independent. Under the law, a person or group is not independent if they were involved in developing, selling, or using the tool, if they have an employment relationship with you or with the tool's vendor during the audit, or if they have a financial interest in either party.
There is no city-run certification and no approved list of auditors. DCWP does not vet, register, or recommend them. That makes choosing a competent and demonstrably independent auditor your responsibility: the law sets the independence bar, but it does not check anyone's qualifications for you.
Publishing the summary of results
Before you use the tool, you have to post a summary of the most recent audit where candidates can find it, which in practice means the careers or jobs section of your website. The summary has to include:
-
The date of the audit and the date you started using the tool,
-
The source of the data used and an explanation of it,
-
The number of applicants or candidates assessed,
-
The selection or scoring rates and the impact ratios for every group, including the intersectional ones,
-
The number of people assessed whose demographic data was unknown.
You can satisfy this with a clearly labeled link to the results. The summary has to stay posted for at least six months after the last time you use the tool.
Notifying candidates
You also have to tell candidates and employees that an AEDT is being used, at least ten business days before it’s used on them. You can give this notice in the job posting, on your website, or directly by mail or email.
The notice has to explain how to request an alternative selection process or an accommodation, if one is available. The law gives candidates the right to ask; it does not require you to offer an alternative. Separately, if a candidate asks, you have to disclose the type of data the tool collects, where it comes from, and how long you keep it.
Penalties and enforcement
DCWP enforces the law, largely in response to complaints, which anyone can file through the city's 311 system or DCWP's website.
The penalties are civil fines. A first violation costs up to $500. Each violation after that runs from $500 to $1,500. They mount quickly, because every day you use a tool without a valid audit is a separate violation, and every failure to give the required notice is its own separate violation. Local Law 144 does not create a new private lawsuit on its own, though candidates may still have claims under other anti-discrimination laws.
Enforcement has been relatively light so far, but that is not a safe reason to skip it. An independent bias audit is increasingly something enterprise customers, partners, and candidates expect to see, and the cost of becoming compliant is small next to the cost of a public fight over a hiring tool.
How to comply, step by step
For most employers, becoming compliant is a short, well-defined project:
-
Identify the tools in scope. List the software you use in hiring and promotion, and decide which ones meet the "substantially assist or replace" test. An auditor can help you make the call.
-
Choose an independent auditor. Confirm they have no involvement with the tool or its vendor and no financial stake in the outcome.
-
Prepare the data. Gather the historical data from your use of the tool, or arrange for test data if you do not have enough.
-
Run the audit. The auditor calculates the selection or scoring rates and impact ratios across every required group and intersection.
-
Publish the summary and post your notices. Put the results on your site, and add the candidate notice to your job postings or application flow at least ten business days before use.
-
Keep it current. Re-audit at least once a year, and keep the summary posted for as long as the tool is in use.
The trend is spreading
New York City was first, but it is no longer alone. New Jersey has proposed a similar audit requirement aimed at tool vendors, Illinois and California have brought AI hiring tools squarely under their anti-discrimination rules, and the European Union treats AI used in hiring as high-risk under its AI Act. The specifics differ, but the direction is consistent: if you use AI to make decisions about people, expect to be asked to show it has been checked. Getting Local Law 144 right is a head start on the rest.
Run your bias audit with Conformance AI
Conformance AI is an independent, third-party AI testing firm, which is exactly what Local Law 144 calls for. We do not build, sell, or operate hiring software, and we have no stake in your results, so the audit we deliver meets the law's independence requirement.
We handle the bias audit end to end. We help you confirm which tools are in scope, prepare and analyze the data, calculate the selection and scoring rates and impact ratios across every required group and intersection, and produce the summary of results in a form you can publish. We can also supply the candidate-notice language to put in front of applicants. You get a defensible audit and the documentation to back it up, without building the expertise in-house.
Because we are a technical AI testing firm rather than a paperwork service, the same audit can go deeper if you ever want it to. For most clients the goal is simpler: meet Local Law 144 cleanly and quickly. That is a good place to start.
If a New York City bias audit is on your list, talk to us.
This article is general information, not legal advice. Local Law 144 and its rules can change; confirm your specific obligations against the current rules and with your own counsel.
