NYC Local Law 144 Checker
Do you use an automated employment decision tool for a New York City job? Work through the AEDT definition in Local Law 144 — the bias audit, the published summary and the 10-business-day notice — and see what likely applies.
Local Law 144 of 2021 added §§20-870 to 20-874 to the New York City Administrative Code. It took effect on 1 January 2023, and the Department of Consumer and Worker Protection began enforcing it on 5 July 2023. The law does not ban automated hiring tools. It wraps one defined thing — the automated employment decision tool — in three duties: an annual independent bias audit, a published summary of that audit, and advance notice to the people the tool is used on. Employers and employment agencies that use an AEDT “in the city” carry all three duties themselves; the vendor who built the tool carries none of them directly.
What this checks
Since 2023, New York City has required employers and employment agencies that screen people with an “automated employment decision tool” — an AEDT — to audit the tool for bias every year, publish the results, and warn candidates before the tool touches their application. Plenty of hiring software is not an AEDT, and plenty of software sold as a simple “helper” is one. The law turns on three elements, and this checker tests each one separately, because failing any single element takes the tool outside the definition.
Everything runs in your browser. Your answers are never sent to a server; they are encoded into a link only if you choose to copy a result link.
- NYC Department of Consumer and Worker Protection — Automated Employment Decision Tools (Local Law 144 of 2021; Administrative Code §§20-870 to 20-874)
- DCWP — AEDT frequently asked questions (the three-element definition and the 6 RCNY §5-300 test for “substantially assist or replace”)
Last verified: 16 November 2026.
Quick answers
What is an AEDT under NYC Local Law 144?
A computer-based tool that meets all three elements in the DCWP definition: it uses machine learning, statistical modelling, data analytics or AI; it produces a simplified output such as a score, classification or recommendation; and that output substantially assists or replaces discretionary decision-making in screening candidates for hiring or employees for promotion. Under 6 RCNY §5-300, “substantially assist or replace” means the output is relied on alone, weighted more than any other single criterion, or used to overrule other conclusions including a human’s.
All three elements must hold at once. A tool missing any one element — plain software with no learned model, rich reports with no simplified output, or a score that is genuinely one light input among many — falls outside the LL144 definition, though other laws may still reach it.
What must an employer do before using an AEDT in New York City?
Three things. Get a bias audit by an independent auditor completed not more than one year before the use. Publish a summary of the audit results, the tool’s distribution date and the audit date on the employment section of your website, and keep it posted for at least six months after the last use. And notify each NYC-resident candidate or employee at least 10 business days before use — that an AEDT will assess them, which qualifications and characteristics it uses, and how to request an alternative process or accommodation.
On request, the employer must also provide the type and source of data the tool collects and the data retention policy, and DCWP expects that notice on the website too. Penalties run up to $500 for a first violation and $500 to $1,500 for each subsequent one, per day of use.
We bought the tool from a vendor. Who is responsible under LL144?
You are. Local Law 144 places the bias audit, publication and notice duties on the employer or employment agency that uses the AEDT in New York City — not on the vendor. The vendor’s job is to make compliance possible: auditor access, the underlying data, and the facts about qualifications, data sources and retention that the notices need.
That split is why procurement questions matter more than contract labels. Before you buy, ask how an independent auditor gets access, what selection-rate data the vendor can produce, and who writes the published summary. If a vendor cannot support the audit, the legal risk stays with you anyway.
How to use
- Question 1: place your role. Vendors and candidates get their own result straight away — the duties sit with employers and agencies, so those two roles need different answers.
- Questions 2 and 4: confirm the position is located in New York City (a NYC-based remote role counts) and that the tool screens people for hiring or promotion.
- Question 3 is the heart of it: answer the three AEDT elements one at a time — the technology, the simplified output, and whether that output substantially assists or replaces human discretion in practice.
- Read your result tier with its duties checklist, check the dates and thresholds table below, and copy the result link if you need to share the answers with counsel.
About this tool
The definition does the real work, and DCWP’s FAQ breaks it into three elements. The tool must be a computational process derived from machine learning, statistical modelling, data analytics or artificial intelligence. It must issue a simplified output — a score, a classification or a recommendation. And that output must substantially assist or replace discretionary decision-making, which the rule at 6 RCNY §5-300 pins down with a practical test: the output is relied on by itself with no other factors, or it sits in a set of criteria weighted more heavily than any other single criterion, or it is used to overrule conclusions drawn from other factors, including a human decision-maker’s. A tool that merely organises applications, or a score recruiters genuinely treat as one light input among many, falls outside. A score that quietly decides who gets interviewed falls inside.
The duties have clocks attached, and the clocks are where employers slip. The bias audit must be completed by an independent auditor not more than one year before the use, and it calculates selection or scoring rates and impact ratios across sex, race and ethnicity, including intersectional categories. The summary of results, the tool’s distribution date and the audit date go on the employment section of the employer’s website and stay there for at least six months after the last use. Each New York City resident candidate or employee gets notice at least ten business days before the tool is used: that an AEDT will assess them, which qualifications and characteristics it uses, and how to request an alternative process or accommodation — plus, on request, the type and source of the data and the retention policy. Penalties under §20-872 run to $500 for a first violation and $500 to $1,500 for each later one, with every day of non-compliant use a separate violation. This checker runs entirely in your browser and stores nothing.
How it works
The mechanism is a definition with three locks, and the law only turns when all three keys turn together. That structure is deliberate: the City did not want to regulate “AI” in the abstract, so it regulated a shape — learned model, simplified output, real decision weight. The transferable signal is to audit the shape, not the label. Vendors call the same product a “decision support insight” in the brochure and a “predictive hiring score” in the sales deck; neither phrase matters. What matters is whether a number comes out and whether that number moves the decision.
Element (c) is where the law quietly becomes a rule about management practice rather than software. Sole reliance, heaviest weighting, overruling — all three describe what people do with the output on a tired Friday afternoon, not what the settings page says. So the honest test is behavioural: pull twenty recent files and ask what actually happened to candidates the tool scored low. If the answer is “none of them got through”, the tool substantially assists your decisions whatever the policy claims. The judgement signal generalises well beyond New York: in any “human in the loop” claim, the loop is real only if the human sometimes wins, visibly, on the record.
The audit duty is a disclosure machine, not a fairness certificate. Local Law 144 never says a tool with a bad impact ratio is illegal to use; it says the ratio must be calculated by an independent auditor and published where candidates can read it. The pressure is reputational and evidentiary — a published summary becomes Exhibit A in any later discrimination claim, and a missing summary is a violation you can prove from a web browser. Read the law that way and the compliance order sorts itself: audit first, publish honestly, notify on time — and treat the numbers you publish as numbers you may one day have to defend.
Worked examples
Resume-ranking platform at a NYC bank: A bank uses a vendor platform that scores and ranks every applicant for New York City roles, and recruiters interview from the top of the ranking. Learned model, simplified output, heaviest weight in practice — all three elements hold. Result pattern: likely covered, full LL144 duties.
Scheduling tool with no score: An employer uses software that books interviews, sends reminders and tracks which stage each candidate sits at. No model ranks anyone and no output feeds the decision. Element (a) and (b) both fail. Result pattern: likely not an AEDT.
Vendor selling into New York: A startup sells a screening tool to NYC employers and asks what it must publish. Nothing — the duties belong to its customers. What it must build is auditor access, selection-rate data and the facts its customers’ notices need. Result pattern: vendor note.
Candidate screened with no warning: A jobseeker learns after rejection that a tool scored her application for a Manhattan role. She was entitled to notice 10 business days before the use, a published audit summary on the employer’s site, and a way to request an alternative process. Result pattern: candidate note.
Quick reference
| Item | Local Law 144 rule |
|---|---|
| Law / citation | Local Law 144 of 2021; NYC Administrative Code §§20-870 to 20-874; rules 6 RCNY §§5-300 to 5-304 |
| In force / enforcement | Took effect 1 January 2023; DCWP enforcement began 5 July 2023 |
| Who carries the duties | Employers and employment agencies using an AEDT in the city — not the vendor |
| Bias audit | Independent auditor; completed not more than 1 year before the use; selection/scoring rates and impact ratios by sex, race/ethnicity and intersectional categories |
| Publication | Summary of results + distribution date + audit date on the employment section of the website; kept posted at least 6 months after the last use |
| Notice | To each NYC-resident candidate/employee at least 10 business days before use: AEDT use, qualifications and characteristics assessed, how to request an alternative process or accommodation; data type/source and retention policy on request and on the website |
| Penalties (§20-872) | Up to $500 first violation; $500 to $1,500 each subsequent violation; each day of use in violation is a separate violation |
Sources: DCWP — Automated Employment Decision Tools; DCWP — AEDT FAQ. Last verified: 16 November 2026.
Frequently asked questions
Is this legal advice?
No. This checker is general information built from Local Law 144, the Administrative Code sections and DCWP’s published guidance and rules. It cannot weigh your hiring process or how DCWP would read your facts. Treat a “likely” result as the start of a conversation with New York employment counsel, not the end of one.
Our recruiters make the final call. Does that keep us outside?
Not by itself. The 6 RCNY §5-300 test looks at the output’s real weight: if the score is the heaviest single criterion, or it is used to overrule what interviews and other factors conclude, the tool substantially assists the decision even though a human signs it off. A human in the loop only helps if the loop is real — the human can and does decide against the score, on the record.
Does a simple keyword filter or spreadsheet count as an AEDT?
Usually not, on the first element. The definition starts with a process derived from machine learning, statistical modelling, data analytics or AI. A hand-written keyword match or a plain spreadsheet formula has no such model behind it. The moment the filter learns from past hiring data, or a vendor’s analytics produce the ranking, element (a) switches on and the other two elements decide.
The role is remote. Is it “in the city”?
It depends on where the position is located, not where the person happens to sit. DCWP guidance treats a position located in New York City as covered even when the work is done remotely. A role based in another state, worked from a Brooklyn apartment, points the other way — get advice on your exact setup, because this is the fact employers most often get backwards.
What happens if we just do not do the audit?
Civil penalties under §20-872: up to $500 for the first violation and $500 to $1,500 for each subsequent violation, and each day the tool is used in violation counts as its own violation. A busy hiring round can stack that up fast. Candidates who never got their 10-business-day notice are also the people most likely to complain to DCWP.
Sources and verification
Built from Local Law 144 of 2021 (NYC Administrative Code §§20-870 to 20-874), the DCWP Automated Employment Decision Tools page and DCWP AEDT FAQ, and the rules at 6 RCNY §§5-300 to 5-304, all on nyc.gov. Last verified: 16 November 2026.
First time embedding? Read the step-by-step guide →
More free tools
EU AI Act High-Risk Checker
Is your AI system high-risk under the EU AI Act? Work through Annex I, Annex III, the Article 6(3) filter and the general-purpose AI tests to see which regime likely applies — and from which date.
Illinois AI in Employment Checker (HB 3773)
From January 1, 2026, Illinois employers must give notice when AI touches hiring and other listed employment decisions — and may not use AI with a discriminatory effect, or ZIP codes as a proxy. Four questions, one narrow law.
Brazil LGPD Checker
Does Brazil’s LGPD cover your processing — and do you qualify as a small processing agent (agente de tratamento de pequeno porte) with simplified duties? Check the Article 3 nexus, the Article 4 exclusions and the Resolução 2/2022 high-risk test.
US State Privacy Checker
Does a US state privacy law cover your business? Pick one state — California, Virginia, Colorado, Connecticut, Texas or Utah — and walk through that state’s own applicability test, threshold by threshold.
Invoice Generator
Create professional invoices in seconds — download as PDF, no sign-up.
Meeting Cost Calculator
See what your meetings really cost — live ticker as time passes.