NYC Local Law 144: What HR Teams Need to Know, Inside New York and Out

New York City’s AI hiring law can reach a company with no headquarters in New York.

The city’s Department of Consumer and Worker Protection (DCWP) ties it to the job’s location. A role counts if it’s based in an NYC office, even part time, or if it’s fully remote and the location associated with it is an NYC office. So a team in Austin or Berlin with two remote roles attached to a small New York office has a question to answer before its screening tool reads another application.

What NYC Local Law 144 regulates

The law is Local Law 144 of 2021. Its text sits in the city’s Administrative Code at §20-870 to §20-874, and the law as enacted is on the City Council’s own site. The statutory definition of an automated employment decision tool (AEDT) runs to one long sentence. DCWP’s FAQ breaks it into three parts that all have to be true. An AEDT is a computer-based tool that:

• “Uses machine learning, statistical modeling, data analytics, or artificial intelligence.”
• “Helps employers and employment agencies make employment decisions.”
• “Substantially assists or replaces discretionary decision-making.”

“Employment decision” reaches further than the final hire. The FAQ says the law “defines employment decision more broadly to include screening.” If an AEDT substantially helps you assess or screen candidates “at any point in the hiring or promotion process,” the requirements apply before you use it.

Whether a specific tool you use meets that definition is a separate question, answered one tool at a time.

The law was enacted in 2021, took effect on 1 January 2023, and has been enforced since 5 July 2023. As of October 2026, we found no amendment or repeal on the city’s official pages.

Does Local Law 144 reach you? The three-part test

DCWP says the law applies only to employers and employment agencies that use an AEDT “in the city,” and then lists what that means. Any one condition is enough. Check each role on its own, because two jobs at the same company can give different answers.

DCWP’s condition	Check it against each role
”The job location is an office in NYC, at least part time.”	Is any part of this role based in an NYC office?
”The job is fully remote but the location associated with it is an office in NYC.”	Is this remote role attached to an NYC office?
”The location of the employment agency using the AEDT is NYC or, if the location of the employment agency is outside NYC, one of the bullets above is true.”	Is a recruiting agency using the tool on your behalf, and where is it?

Your headquarters appears nowhere in that table. A company based in Denver whose only New York presence is a small sales office has two kinds of role to check: the ones based in that office, and the remote ones attached to it.

The third row catches a setup that’s easy to miss. If an agency screens candidates for you with its own tool, the agency’s location matters too.

None of the three conditions names where a candidate lives. Residence comes up later, in who receives notice.

Every row also assumes the tool is an AEDT used for an employment decision. If you’re mapping where AI already touches your hiring, the 11 agents across talent acquisition set out what each one does and who decides.

Run this against each open role, then confirm the answer with your employment counsel.

What the law requires once it applies

Three duties, each with its own clock.

Duty	What it requires	Timing
Bias audit	An independent auditor works out each group’s selection rate (the share of that group the tool moves forward or puts in a given category) and its impact ratio (that rate divided by the rate of the most-selected group). For a tool that scores people, the auditor uses scoring rates instead (the share of each group scoring above the median) and compares each group with the highest-scoring group. Groups are sex, race/ethnicity, and sex and race/ethnicity combined	Conducted “no more than one year prior to the use” of the tool
Public posting	The date of the most recent bias audit, a summary of its results, and the tool’s distribution date (the date you began using it), on “the employment section of their website in a clear and conspicuous manner”	Before use, and kept posted “for at least 6 months after its latest use of the AEDT for an employment decision”
Candidate notice	Notice, to each candidate or employee “that resides in the city,” that an AEDT will be used to assess them, with a way to request an alternative selection process or accommodation. A second notice names “the job qualifications and characteristics” the tool will use. Information on the data the tool collects, its source and your retention policy must be on your website or available on written request	”No less than ten business days before such use” (both notices)

The audit and notice timing come from the law itself, §20-871. The calculations, the posting rules and the auditor test come from DCWP’s final rules, in force since 5 July 2023.

Those rules are specific about who can audit. An auditor doesn’t count as independent if they are or were “involved in using, developing, or distributing the AEDT.” The same goes for anyone who, during the audit, works for you or the tool’s vendor, or holds “a direct financial interest or a material indirect financial interest” in either.

Penalties are set in §20-872. Anyone who breaks the law or its rules “is liable for a civil penalty of not more than $500 for a first violation and each additional violation occurring on the same day as the first violation, and not less than $500 nor more than $1,500 for each subsequent violation.” For the audit and posting duty in §20-871(a), the statute adds that “each day on which an automated employment decision tool is used in violation of this section shall give rise to a separate violation.” A screening tool that runs every working day without a current audit runs up a new violation each of those days. Each failure to give a required notice is also “a separate violation.”

Other AI hiring rules in Colorado, California and the EU, as of October 2026

If you hire across US states or in the EU, NYC is one rulebook of several. Here are three more, each read on its own government’s site. Check the rules in every state where you hire.

Jurisdiction	Law	Status as of October 2026	Core duty
New York City	Local Law 144	In effect 1 January 2023; enforced since 5 July 2023	Bias audit, public posting, candidate notice (above)
Colorado	SB26-189, “Automated Decision-Making Technology”	Signed by the Governor 14 May 2026; developer duty starts 1 January 2027	The developer of a covered tool gives the deployer (the business using it, such as an employer) technical documentation of its intended uses, categories of training data, known limitations, and instructions for appropriate use and human review
California	Civil Rights Council regulations on automated-decision systems in employment	In effect since 1 October 2025	Employment records, “including automated-decision data,” kept “for a minimum of four years”
European Union	AI Act, Annex III (employment listed as high-risk)	Annex III rules apply from 2 December 2027	An employer using a high-risk AI tool at work informs affected employees and workers’ representatives beforehand

Older articles still cite Colorado’s earlier bill. The legislature’s own page for SB26-189 says that in 2024 it “enacted Senate Bill 24-205,” and that the new act “repeals and reenacts those provisions with new requirements regarding the use of automated decision-making technology in consequential decisions.” If a vendor’s compliance page cites SB24-205, ask which version it was written against.

California’s rules were approved on 27 June 2025, according to the Civil Rights Department. In the department’s words, they “clarify the application of existing antidiscrimination laws in the workplace in the context of new and emerging technologies, like artificial intelligence.”

The EU’s AI Act FAQ gives employment examples: “to place targeted job advertisements, to analyse and filter job applications, and to evaluate candidates.” The duty to inform “goes in addition to any workers’ consultation rules that may already apply.” The AI Omnibus, an EU law that extended the AI Act’s timelines, entered into force on 27 July 2026, the European Commission says. Under it, the Annex III rules apply starting 2 December 2027.

Who decides in a screening agent whose guardrails name this law

On our Agent Map, the Application Screening Support Agent carries the verdict “human-decides.” It reads each application against the published must-have criteria, summarizes the evidence for and against each one, and orders the pool for recruiter review, with reasons a recruiter can check and overturn. The recruiter or hiring manager decides who advances and who’s declined. The agent never auto-rejects.

Its guardrails name this law’s duties directly. Candidate notice goes out before use. The NYC LL144 bias audit and notice, the EU AI Act’s high-risk obligations, and works-council consultation in the EU are all on the list. The agent reads only the published criteria and draws nothing from a name, an address, a school, dates or photos. Demographic data is used only for the aggregate audit, with a minimum group size, and never in screening. Every summary and ordering is logged for audit and can be overturned by the recruiter.

Read that against the duties above. The audit needs selection rates by group. The agent logs every summary and ordering it produces, which gives an auditor a record of what the tool did with each application. The notice goes out before the tool reads anything. The guardrail says “annual bias audit.” The statute ties the audit to each use: it must be “no more than one year prior to the use,” so a tool in daily use needs its next audit before the last one turns a year old.

Two problems sit outside anything a guardrail fixes. Criteria that were never published can’t be screened against, so publish first, then screen. And recruiters read only the top of the queue unless the tool makes them see the whole pool. The design puts every application in front of a person. Whether that person reads past the first page is a habit, and habits are harder to audit than logs.

A screening decision is a decision about a person, in a regulated process, and the framework explains why a decision like this needs real trust to earn before anything is automated. Our post on prioritizing HR AI use cases scored this exact agent end to end. The other 121 agents, and who decides on each, are on the full Agent Map.

---

If AI already touches your hiring, start with a list. Where does it shape a decision about a candidate? Who owns each of those decisions? Where do the guardrails need to sit before anything goes live?

That’s part of what our one-month strategy engagement produces: your decisions, their owners and the guardrails each one needs, written down so your counsel has a clear list to check. The engagement is a fixed fee. Each use case after it is a fixed fee for agreed scope, quoted after the proof.

Book a 30-minute call and bring the screening tool you’re least sure about.

Frequently asked questions