New York City has regulated AI in hiring since 2023, and for its first two years the city enforced the law mostly by waiting for complaints. It received two, between July 2023 and June 2025. When New York's state comptroller audited that record in December 2025, its auditors went through the same companies the city had reviewed and found at least 17 possible violations where the city had found one. The law hasn't changed. The scrutiny has.
What is NYC Local Law 144?
NYC Local Law 144 says that an employer or employment agency can't use an automated tool to screen candidates or employees for a New York City job or promotion unless the tool has had an independent bias audit in the past year, a summary of that audit is on the employer's website, and every candidate or employee who lives in the city was told about the tool at least 10 business days before it was used. The city calls these tools automated employment decision tools, or AEDTs. The law took effect on January 1, 2023, and the Department of Consumer and Worker Protection (DCWP) began enforcing it on July 5, 2023.
This guide is for HR, recruiting, IT and operations leads at companies that hire in New York City. It's based on the law, DCWP's rule and DCWP's own FAQ, and the state comptroller's audit. It isn't legal advice.
Does Local Law 144 apply to your company?
Two questions decide it: is the tool used "in the city", and is it an AEDT?
Is the job in New York City?
There's no size threshold. According to DCWP's guidance, the law applies when:
- the job is based in a New York City office, at least part time; or
- the job is fully remote, but the location associated with it is a New York City office; or
- an employment agency using the tool is located in the city (or, if it's outside the city, one of the first two applies).
The bias audit and posting duties apply to every covered use. The notice duty is narrower: it's owed to candidates and employees who live in the city.
The law also stops at people who have applied. DCWP's FAQ says the requirements don't apply when a tool scans a résumé bank, contacts potential candidates or invites applications. A candidate is someone who has applied for a specific position in the format you asked for. Employees being considered for promotion are covered too.
Is your tool an automated employment decision tool?
The law's definition is broad: any 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". DCWP's rule then narrows what "substantially assist or replace" means. A tool counts only if you:
- rely solely on its output, with no other factors considered; or
- use its output as one of several criteria, weighted more than any other criterion; or
- use its output to overrule conclusions from other factors, including human decision-making.
The rule's examples of simplified output are a score (rating a candidate's estimated technical skills), a tag (categorizing a résumé by keywords), a recommendation (whether to interview someone) and a ranking (ordering candidates by how well their cover letters match the job). Tools that only translate or transcribe, such as converting a résumé from a PDF or transcribing an interview, don't produce simplified output. Junk email filters, firewalls, calculators and spreadsheets are excluded in the law itself.
That makes coverage depend as much on your process as on the product. The same applicant tracking system can be in scope at one company and out of scope at another, depending on whether recruiters work from its match scores or look past them.
Likely covered:
- Your applicant tracking system scores or ranks applicants, and recruiters call the top of the list.
- An assessment or video-interview platform scores candidates, and a cut-off score decides who moves on.
- A recruiter pastes résumés into a chatbot, asks it for a shortlist, and interviews the shortlist.
- A promotion tool's rating outweighs the manager's view.
Likely not covered:
- AI drafts the job ad, interview questions or rejection emails.
- AI sources or contacts people who haven't applied.
- AI transcribes interviews for a person to review.
- A score is one of several factors and doesn't carry the most weight, and you can show that.
What does Local Law 144 require?
Three things, before you use the tool, and one more on request.
1. A bias audit from the past year
The audit must be done by an independent auditor no more than one year before you use the tool, and you can't keep using the tool once a year has passed since the latest audit. An auditor isn't independent if they were involved in using, developing or distributing the tool, or if, during the audit, they work for you or the vendor or have a financial interest in either. DCWP doesn't approve auditors.
At a minimum, the audit calculates for each category:
- Selection rate, for tools that select or classify: the share of people in a category who move forward. DCWP's example: if 100 Hispanic women apply and 40 are selected for interview, their selection rate is 40%.
- Scoring rate, for tools that score: the share of people in a category who score above the median for the whole sample.
- Impact ratio: a category's rate divided by the rate of the most selected (or highest-scoring) category.
The categories are sex, race and ethnicity, and the intersections of the two, following the federal EEO-1 categories. The audit must also report how many people fell into an unknown category. An auditor may leave out a category that makes up less than 2% of the data, but must explain why and still report its numbers.
Whose data? The audit must use historical data, meaning data from real use of the tool. It can pool data from several employers that use the same tool. You can only rely on a pooled audit if you gave the auditor your own historical data or you've never used the tool before. If there isn't enough historical data for a statistically significant audit, the audit can use test data, and its summary must say why and how the test data was produced. Demographic data inferred by software or imputed can't be used.
What the results oblige you to do. Nothing, under Local Law 144 itself. DCWP's FAQ says the law "does not require any specific actions based on the results of a bias audit", then reminds employers that federal, state and city anti-discrimination laws still apply. A low impact ratio is a published number you'd have to answer for under those laws.
2. A public summary on your careers page
Before you use the tool, post on the employment section of your website, "in a clear and conspicuous manner":
- the date of the most recent bias audit;
- a summary of its results: the source and explanation of the data, the number of people in the unknown category, and the number of applicants, the selection or scoring rates and the impact ratios for every category;
- the tool's distribution date, meaning the date you began using it.
A clearly labeled link to a page with the same information also counts. Keep it posted for at least 6 months after you last use the tool.
3. A notice 10 business days before use
Tell each candidate or employee who lives in the city, at least 10 business days before the tool is used on them:
- that an AEDT will be used to assess or evaluate them;
- the job qualifications and characteristics it will use;
- how to request an alternative selection process or a reasonable accommodation under other laws.
The rule doesn't require you to offer an alternative process. It requires you to say how to ask.
For applicants, you can give notice on the employment section of your website, in the job posting, or by mail or email. For promotions, you can use a written policy given to employees, a job posting, or mail or email. DCWP's FAQ says a website notice doesn't have to be specific to one position, and that you can start using the tool 10 business days after posting it, whenever the job itself was posted.
4. Data information on request
Post on your careers page the type of data the tool collects, its source and your data retention policy for it, plus instructions for asking for that information in writing. If someone asks, you have 30 days to answer. The only exception is where disclosure would break a law or interfere with a law enforcement investigation, and then you must explain why.
Who is responsible: the employer or the vendor?
You are. DCWP's FAQ is direct: employers and employment agencies are responsible for making sure they don't use a tool without a bias audit, and "the vendor that created the AEDT is not responsible for a bias audit of the tool." A vendor may commission an independent audit of its own tool, and many do. That audit only helps you if it meets the pooled-data rule above and the auditor really is independent.
That second point is where the state's auditors found problems. Of the companies they reviewed, three relied on audits not done by an independent auditor. In two of those, an independent auditor wrote the report, but the vendor selling the tool did the actual testing. When you receive a vendor's audit, ask who ran the numbers, not just who signed the report.
What are the penalties, and who enforces them?
DCWP enforces the law. The 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 violation is a separate violation, and each missing notice is a separate violation too. The city's lawyers can also go to court to force compliance. The law doesn't limit a candidate's or employee's right to sue, or the NYC Commission on Human Rights' power to enforce the city's Human Rights Law. DCWP sends discrimination claims about these tools to the Commission.
What the state audit found
The New York State Comptroller published its audit of DCWP's enforcement on December 2, 2025, covering July 2023 to June 2025. Its main findings:
- Two complaints. DCWP received two complaints about these tools in two years, and didn't check whether its intake process worked.
- Calls went elsewhere. In 12 test calls to 311 about filing a complaint, auditors were transferred to DCWP only three times. Eight times they were referred to the New York State Department of Labor.
- Missed problems. DCWP surveyed 32 companies and found one issue in the 22 bias audits it reviewed. The state's auditors reviewed the same companies and found at least 17 instances of potential non-compliance. These included audits that didn't properly calculate selection rates and impact ratios, and audits that didn't properly explain their use of historical data.
DCWP disputed parts of the audit, but agreed it could adopt the recommendations on handling and routing complaints, and said it would review its protocols. The comptroller also recommended that DCWP look for violations through research, not just complaints. Two years of quiet enforcement aren't a reason to assume the next two will be quiet. A summary is posted on your own website, which makes non-compliance easy to find.
A bias audit checks the tool, not the way you use it
The audit measures one thing: whether the tool's outputs differ across groups of people. It says nothing about whether a recruiter looked past the ranking, whether the cut-off score made sense for this role, or whether anyone could have overruled the result. Yet those are exactly the questions that decide whether the law applies to you at all, because the rule's three tests are about how much weight your people give the output.
So the useful work is in your process, not only in the vendor's PDF. Decide, for each tool, what role its score plays and who can override it, and write that down. Shared and stakes, in module 6 of our free course, uses a recruiter ranking résumés as its example of a high-stakes use nobody formally built, and shows how to name the check each one needs. Our framework asks the same of every high-stakes use: the person checking must have the time, the information and the authority to reject the output.
How to comply with Local Law 144
There are six steps, in order. Each produces what the next one needs.
01 · List every tool that touches a hiring or promotion decision
Include your applicant tracking system's scoring, matching and ranking features, assessment and video-interview platforms, and promotion or performance tools. Then ask recruiters and hiring managers what they use chatbots for. The free Shadow AI Audit catches the AI features vendors have switched on inside software you already pay for. Your AI list, in module 3 of our course, turns what you find into a list with an owner on every row.
02 · Apply the three tests and write down the answer
For each tool used for New York City jobs, ask: do we rely on its output alone, weight it above every other criterion, or let it overrule a person? If any answer is yes, it's covered. If every answer is no, write down why, with how the output is actually used, so you can show your reasoning if DCWP asks. Put the rule for chatbots in your AI usage policy.
03 · Get a bias audit you can rely on
Ask each vendor for its latest bias audit and check three things: it's less than a year old on the day you'll use the tool, the testing was done by someone independent of the vendor, and either you supplied your historical data or this is your first use. If any of those fails, commission your own audit. Put the audit date in a calendar with a reminder well before the year is up.
04 · Publish the summary and the distribution date
Post the audit date, the summary of results and the date you began using the tool in the employment section of your website, or link to them with a clearly labeled link. Keep them up for 6 months after you stop using the tool.
05 · Post the notice and the data information
Add a notice to your careers page or job postings: that you use the tool, the qualifications and characteristics it assesses, and how to ask for an alternative process or an accommodation. Add the type and source of data the tool uses, your retention policy for it, and how to request that information in writing. Then wait 10 business days before using the tool. For promotions, put the notice in a written policy you give to employees.
06 · Give each tool an owner and a review date
Name an owner and a backup for each covered tool. Their jobs: renew the audit before it lapses, update the summary when it changes, answer data requests within 30 days, and recheck the three tests whenever the vendor ships a new scoring feature or your recruiters change how they use it. Owners and backups, in module 7 of our course, sets these up. If you'd rather have help with the whole review, our AI governance work covers it.
What's still open
- How DCWP enforces from here. The comptroller recommended that DCWP find violations through research, not only complaints. As of November 2026, DCWP's page on the law describes no change to its approach since the audit.
- A study of the city's own tools. Local Law 25 of 2026 orders the city's technology department to study how algorithmic tools and AEDTs affect the city's own employees and report on it. It doesn't change the duties on private employers.
- Other laws on top. If you also hire in states with their own AI hiring laws, such as Colorado or Connecticut, their duties apply alongside this one.
We'll update this article when the law, DCWP's rule or its enforcement approach changes. The date at the top of the page shows when we last checked.
Sources
- Int 1894-2020, Local Law 144 of 2021, New York City Council, 2021: the law's text (Administrative Code §§ 20-870 to 20-874), including the definitions, notices, penalties and effective date.
- Notice of Adoption of Final Rule on Automated Employment Decision Tools, NYC Department of Consumer and Worker Protection, April 2023: the rule (6 RCNY §§ 5-300 to 5-304), with the three tests, the audit calculations and the worked examples.
- Automated Employment Decision Tools: Frequently Asked Questions, DCWP, June 2023: where the law applies, who's responsible for audits, notice options and complaints.
- Automated Employment Decision Tools (AEDT), DCWP: the enforcement start date and how to file a complaint.
- Enforcement of Local Law 144: Automated Employment Decision Tools, Office of the New York State Comptroller, December 2025: the audit's summary, and the full report 2024-N-6 with the test calls, the bias audit review and DCWP's response.
- Local Law 25 of 2026, New York City Council, January 2026: the study of algorithmic tools and AEDTs used on city employees.