NYC Local Law 144: Bias Audits, Notices and Compliance

New York City's law requires employers to have their hiring software bias-audited, publish the results and tell candidates before the tool is used. If you hire for any role connected to a New York City office, the law probably reaches at least one tool in your stack.

This guide sets out what the law covers and what you have to do, with a checklist at the end. It is a plain-English explainer drawn from the law, the rules of the city's Department of Consumer and Worker Protection (DCWP) and DCWP's FAQ, as of October 2026. It is not legal advice, so confirm how it applies to you with counsel. For how New York compares with the EU, Colorado and Illinois, see our guide to AI hiring laws in 2026.

What is NYC Local Law 144?

NYC Local Law 144 of 2021 prohibits employers and employment agencies from using an automated employment decision tool in New York City unless a bias audit was done and the required notices were given. It was enacted on 11 December 2021, took effect on 1 January 2023, and DCWP began enforcing it, with its implementing rules, on 5 July 2023.

The law sits in the city's Administrative Code (sections 20-870 to 20-874). DCWP's rules, in title 6 of the Rules of the City of New York, fill in the detail: what counts as a covered tool, how the audit is calculated and what you must publish.

What counts as an automated employment decision tool?

An AEDT is a computer-based tool that uses machine learning, statistical modelling, data analytics or artificial intelligence, produces a simplified output such as a score, tag, classification, recommendation or ranking, and substantially assists or replaces discretionary decisions about hiring or promotion. All three parts have to be true.

The rules give "substantially assist or replace" a narrow, specific meaning. A tool is in scope if you:

1. rely solely on its output, with no other factors considered; or
2. use its output as one of several criteria, but weight it more than any other criterion; or
3. use its output to overrule conclusions drawn from other factors, including a person's judgement.

In practice that can bring in résumé-ranking and matching tools, automatically scored assessments and video interviews, and chatbots that decide who moves forward, whenever their output carries that much weight. The rules exclude tools that only translate or transcribe, such as converting a CV from PDF or transcribing an interview. The law itself excludes everyday software that does not support decisions, such as spam filters, firewalls, antivirus software, calculators, spreadsheets and databases.

Screening at any stage counts, so a tool that only decides who gets a first interview is covered even if a person makes the final hire. Equally, if a recruiter reviews every candidate and the tool's score is one factor weighted no more heavily than the others, the tool may fall outside the definition. That depends on how your team actually uses the output, so write it down.

Who does the NYC bias audit law apply to?

It applies to employers and employment agencies that use an AEDT "in the city". DCWP's FAQ says that means the job is based at a New York City office at least part of the time, or the job is fully remote but tied to a New York City office, or the employment agency using the tool is located in the city.

The notice duty runs to candidates and employees who live in New York City. The law also covers promotions as well as hiring, so internal mobility tools are in scope on the same terms.

There is no size threshold. A ten-person firm with a New York office has the same duties as a large one.

What the law does not cover matters too. DCWP's FAQ says the audit and notice requirements apply only to people who have applied for a specific position, or employees being considered for promotion. Using software to search a CV database, run outreach or invite people to apply is outside those requirements. Sourcing tools are therefore usually out of scope, while screening tools usually are not. Our overview of AI recruitment tools shows where each category sits in a typical process.

What does an AEDT bias audit include?

An AEDT bias audit is an impartial evaluation by an independent auditor that calculates, at a minimum, selection or scoring rates and impact ratios for sex categories, race/ethnicity categories and intersectional categories of the two. The categories are the EEO-1 Component 1 categories US employers already use for federal reporting.

How the numbers are calculated depends on what the tool does.

A worked example, with hypothetical figures: if 200 women and 200 men apply and the tool advances 50 women and 80 men, the selection rates are 25% and 40%. The most selected category is men, so the impact ratio for women is 25 / 40 = 0.625. The rules' own example uses the same arithmetic: 40 of 100 Hispanic women selected for interview is a 40% selection rate.

The audit must also:

• run the calculations separately for each group if the tool sorts people into groups, such as leadership styles;
• report how many people the tool assessed whose sex or race/ethnicity is unknown, since they are left out of the calculations;
• justify any category excluded for making up less than 2% of the data, and still report its numbers.

What data can the audit use? Historical data from real use of the tool is the default. An audit can pool data from several employers that use the same tool, which is how a vendor-commissioned audit typically works, but you can rely on a pooled audit only if you supplied your own data to the auditor or you have never used the tool before. If there is not enough historical data for a statistically significant audit, the auditor can use test data, and the summary must explain why and how it was produced. DCWP's FAQ says demographic data cannot be imputed or inferred by software.

Who counts as independent? Someone who can exercise objective, impartial judgement. An auditor is not independent if they were involved in using, building or distributing the tool, work for you or the vendor during the audit, or have a direct or material indirect financial interest in either. DCWP keeps no list of approved auditors.

Does a low impact ratio mean you must stop? The law sets no pass mark and requires no action based on the results. Federal, state and city anti-discrimination laws still apply, though, and the federal Uniform Guidelines generally treat a selection rate below four-fifths (80%) of the highest group's rate as evidence of adverse impact, with caveats for small samples. A poor ratio in a published audit is evidence a claimant can use. Our guide to unconscious bias in recruitment covers the human side of the same problem.

Who is responsible for the bias audit: the employer or the vendor?

The employer or employment agency using the tool is responsible. DCWP's FAQ says the vendor that created the tool is not responsible for a bias audit, though a vendor may have an independent auditor carry one out. Employers remain responsible for making sure an audit was done before use.

You carry the risk if the vendor's audit is out of date, does not cover your use, or was done by someone who is not independent. Ask for the full summary, check the date and diarise the renewal. Our checklist for choosing an assessment platform includes the audit questions to put to vendors, and the enterprise platforms guide covers adverse-impact questions for larger RFPs.

What must you publish, and for how long?

Before you use the tool, you must publish on the employment section of your website, clearly and conspicuously, the date of the most recent bias audit, a summary of its results and the tool's distribution date, meaning the date you started using it. A clearly labelled link to a page holding the summary, such as the vendor's, also satisfies the rule.

The summary has to include:

• the source of the data used and an explanation of it;
• the number of people assessed who fall in an unknown category;
• the number of applicants or candidates, the selection or scoring rates, and the impact ratios for every category.

Keep the summary and distribution date posted for at least six months after you last use the tool for an employment decision. Retiring a tool does not end the duty straight away.

What notice do candidates have to receive?

Candidates and employees who live in New York City must be told, at least 10 business days before the tool is used, that an AEDT will be used and which job qualifications and characteristics it will assess. The notice must include instructions for requesting an alternative selection process or a reasonable accommodation under other laws.

You can give notice to job applicants in any of three ways:

For employees considered for promotion, a written policy or procedure given to employees at least 10 business days ahead is a further option. The careers-site notice covers every role at once, so you do not have to wait 10 business days after each new posting.

The rules do not oblige you to provide an alternative selection process, only to explain how to request one. Disability law may still require an accommodation, so have a route ready and a named owner for requests. A structured interview scored on a shared scorecard is a defensible alternative for most roles.

Data disclosure. You must also post, on the employment section of your site, your AEDT data retention policy, the type of data collected and its source, plus instructions for requesting it in writing. Answer written requests within 30 days, unless disclosure would break the law or interfere with a law enforcement investigation, in which case explain why.

How is NYC Local Law 144 enforced, and what are the penalties?

DCWP enforces the law, mostly through complaints. Anyone can report an employer that used an AEDT without an audit, without posting the summary or without giving notice, through 311 or DCWP's online complaint form, and no account is needed. DCWP refers claims of discrimination to the NYC Commission on Human Rights, which enforces the city's Human Rights Law.

The civil penalties are:

• up to $500 for a first violation, and for each additional violation on the same day;
• $500 to $1,500 for each subsequent violation;
• each day a tool is used in breach counts as a separate violation;
• each failure to give a required notice counts as a separate violation.

The city's Corporation Counsel can also go to court to force compliance. The law does not limit candidates' rights to sue under other laws, so a published audit can feature in a discrimination claim brought under federal, state or city law.

NYC Local Law 144 compliance checklist

1. Inventory every hiring and promotion tool. For each, note what it outputs (score, rank, tag, pass or fail) and who acts on it.
2. Test each against the three meanings of "substantially assist or replace". Record the reasoning, especially for tools you decide are out of scope.
3. Map which roles are "in the city". Include remote roles tied to a New York City office and work placed through New York agencies.
4. Get the audit. Obtain the vendor's most recent bias audit summary, check it is dated within the last year and the auditor is independent, and send the auditor your own historical data if you want to rely on a pooled audit after your first year of use.
5. Publish. Post the audit date, summary and distribution date on your careers site, or a clearly labelled link to them, before first use.
6. Post the notice. Put a general AEDT notice on your careers site, listing what each tool assesses and how to request an alternative or accommodation. Wait 10 business days before using the tool.
7. Post the data disclosure. Retention policy, data types and sources, and how to make a written request. Assign someone to answer within 30 days.
8. Set up the alternative route. Decide what happens when a candidate asks, and who owns it.
9. Calendar the renewal. Diarise the next audit at least two months before the current one expires, and review the impact ratios with HR and counsel when it arrives.
10. Keep it posted. When you retire a tool, leave its summary up for six months after last use.

Where does Cohesyve fit?

Cohesyve takes a job description, sources candidates, runs outreach, puts candidates through AI job simulations of the real work and returns a ranked shortlist. A ranked shortlist is the kind of simplified output Local Law 144 describes. If you hire for New York City roles, treat Cohesyve like any other screening tool: put it on your inventory, cover it in your notice, and talk to us about audit data before using it for those roles. Plans start at $199 a month billed annually, and new accounts include 10 free candidates.

Common questions

Does NYC Local Law 144 apply to companies based outside New York? Yes, if the job is in the city. The test is where the job is located, or which office a remote role is tied to. Headquarters location does not matter.

Does a bias audit apply if a person makes the final hiring decision? Often, yes. The law covers screening at any stage, so a tool that decides who reaches the interview is covered even when a person makes the offer. What matters is how heavily its output is weighted.

Can we use our vendor's bias audit? Yes. Before your first use you can rely on a vendor audit built on other employers' data. After that, you can keep relying on a pooled audit only if you supplied your own historical data to the auditor.

Is a skills test covered? It depends on how the score is produced and used. The definition needs a computer, at least in part, to decide which inputs matter and how much, so a test marked against a fixed answer key may fall outside it. An assessment scored by AI and weighted above other criteria usually falls inside.

Do we need a separate notice for every job? No. A clear notice on the employment section of your website covers all roles, and you can start using the tool 10 business days after posting it.