NYC Local Law 144 and AI Recruiting Tools — What Agencies Need to Know
NYC Local Law 144, in effect since July 2023, was the first major US regulation of AI-driven hiring tools. It applies to anyone using "automated employment decision tools" (AEDTs) for hiring or promotion of NYC employees or candidates.
If you are a recruiting agency using AI matching, semantic search, or automated candidate ranking, this law might apply to you. This post breaks down exactly when and how — and where the boundaries are.
This is not legal advice. Talk to an employment lawyer for your specific situation.
What LL 144 actually says
The text of NYC Local Law 144 (and the implementing rules from the NYC Department of Consumer and Worker Protection) requires that anyone who uses an AEDT to substantially assist or replace discretionary decision-making in hiring or promotion of an NYC-located candidate must:
- Conduct a bias audit of the AEDT in the prior calendar year.
- Publicly post the bias audit results on the employer / agency's website.
- Notify candidates that an AEDT will be used at least 10 business days before use.
- Allow candidates to request an alternative selection process (in some readings; the rule is vague here).
- Identify the data sources the AEDT uses, on request.
Failure to comply triggers civil penalties from $500 to $1,500 per violation per day.
What is an "automated employment decision tool"?
This is the operative question and the answer is narrower than most people think.
The DCWP rules define an AEDT as a tool that uses machine learning, statistical modeling, data analytics, or AI that issues a "simplified output" (score, ranking, classification, recommendation) that "substantially assists or replaces discretionary decision-making."
"Substantially assists or replaces" is defined as one of:
- The output is the only criterion for the decision, OR
- The output is given more weight than other inputs the human considers, OR
- The output overrides any human discretionary input.
In plain English: if the AI score determines who gets the interview, you are using an AEDT. If the AI score is one of many inputs and the recruiter overrides it freely, you probably are not.
Does LL 144 apply to recruiting agencies?
This is the big question. The law covers "employers" and "employment agencies." Recruiting agencies clearly count as employment agencies for NYC roles.
Where it definitely applies:
- Your tool ranks candidates and the top 5 are the only ones reviewed by the recruiter.
- Your tool generates a "match score" and candidates below a threshold are auto-rejected.
- Your tool screens cover letters and shortlists candidates without human review.
Where it probably does not apply:
- Your tool is purely for sourcing (finds candidates from the open web, but does not rank or filter them).
- Your tool generates suggestions that a recruiter independently reviews and frequently overrides.
- Your tool is used to organize data but does not output rankings used for selection.
The phrase that matters: "substantially assists." If the human recruiter is making the final call with full information, you are almost certainly outside scope. If the algorithm is the gatekeeper, you are inside.
Common recruiting tool features and their LL 144 status
| Feature | Likely AEDT? | Why |
|---|---|---|
| Boolean keyword search | No | No ML or statistical modeling |
| Semantic candidate search (SBERT, FAISS) | Probably no | Returns ranked candidates but recruiter reviews each |
| Auto-screening (resume → score → reject) | Yes | Algorithmic gatekeeping |
| AI generated personalized outreach | No | Affects communication, not selection |
| AI generated candidate rankings shown to recruiter | Maybe | Depends on recruiter override frequency |
| Predictive job-candidate fit score | Maybe | Depends on whether it gates or assists |
| Resume parsing and field extraction | No | Data normalization, not selection |
| Diversity-balanced shortlisting | Yes | Algorithm changes selection |
The maybes are where most agencies need to think carefully.
What a bias audit actually looks like
A LL 144-compliant bias audit must:
- Be conducted by an independent auditor (not the vendor of the AEDT, not in-house).
- Calculate selection rates and impact ratios across race, ethnicity, and sex.
- Be conducted annually.
- Cover at least the prior calendar year of use data.
Typical cost: $5,000 to $25,000 per AEDT depending on data complexity.
The most important number the audit produces: the impact ratio for each protected category. This is the selection rate for the group divided by the selection rate for the highest-selected group. A ratio under 0.8 (the four-fifths rule) signals adverse impact.
What goes on your public website
A LL 144 audit posting must include:
- Date of the most recent audit
- The summary of audit results (selection rates and impact ratios across race, ethnicity, sex)
- The category of AEDT (e.g., "applicant scoring system")
- The distribution date when the AEDT was first used
Most agencies publish this on a /ai-audit or /compliance page on their website. Keep it updated annually.
Candidate notification requirements
If you use an AEDT for an NYC role, you must notify candidates:
- At least 10 business days before the AEDT is used.
- Disclose what job qualifications and characteristics the tool will assess.
- Allow alternative selection process requests.
Practical implementation: add a paragraph to job postings or candidate-facing communications. Wording that has held up:
"We use automated tools to assist in evaluating candidate qualifications including {list of qualifications}. If you would like to request an alternative selection process, contact {recruiter_email} at least 10 business days before your scheduled assessment."
What happens if you ignore LL 144
The DCWP can issue civil penalties:
- $500 for the first violation
- $500 to $1,500 per subsequent violation
- Each day of non-compliance is a separate violation
Plus reputational risk if a high-profile candidate complaint goes public, plus potential exposure in any disparate-impact litigation that follows from a complaint.
So far enforcement has been moderate. We are aware of fewer than 20 actions in the first year of the law. But that will grow as the DCWP staffs up.
What other jurisdictions are doing
LL 144 was the first. It will not be the last:
- Illinois Artificial Intelligence Video Interview Act (2020): regulates AI analysis of video interviews.
- California AB 331 (proposed, multiple times): broader AI hiring regulations.
- Colorado SB 21-169: anti-discrimination requirements for AI in insurance, with hiring implications.
- EU AI Act: classifies hiring AI as "high-risk" and imposes broad requirements as of February 2025.
- NYC LL 145 (potential follow-up): expansion of LL 144 scope.
If you operate in NYC, plan for LL 144 to become the floor and other jurisdictions to ratchet up.
How placement.solutions handles LL 144
A few design choices we made specifically with LL 144 in mind:
- Match scores are advisory, not gatekeeping. We show recruiters a ranked candidate list but the full database is always one click away. We do not auto-reject anyone.
- Recruiter override is frictionless. Recruiters can advance any candidate at any stage, regardless of match score, with one click.
- Audit data export: customers using AEDT-type features can export selection data for their own bias audit.
- Per-tool documentation: every AI feature has plain-English documentation of what it does and what data it uses.
In our reading, the way most agencies use placement.solutions falls outside the AEDT definition because the recruiter retains full discretion. But customers who use heavy auto-screening features should run a bias audit and follow the disclosure requirements.
If you are unsure where your use of any tool falls, talk to an employment lawyer who specializes in AI compliance. The few thousand dollars in legal fees are cheap relative to civil penalties or litigation exposure.
A practical compliance checklist
If you are a recruiting agency operating in NYC and using any AI hiring tool, run this checklist annually:
| Item | Status |
|---|---|
| Inventory all AI tools used in candidate selection | |
| Determine which (if any) qualify as AEDTs | |
| Run bias audit on each AEDT | |
| Post audit results on public website | |
| Add candidate notification language to job postings | |
| Implement alternative-selection-process workflow | |
| Document data sources for each AEDT | |
| Train recruiters on disclosure obligations | |
| Set annual review reminder |
Where the law is heading
Expect three trends in 2026 and beyond:
- More aggressive enforcement as the DCWP staffs up.
- Class action attempts from candidates claiming AEDT-driven adverse impact.
- Vendor liability shift — agencies will start demanding indemnification from AEDT vendors.
If you are a vendor (we are), this means more rigorous documentation and audit cooperation. If you are an agency, this means careful tool selection and clear documentation of how you use each tool.
About placement.solutions: Built for recruiting agencies, with LL 144 considerations baked into the design. Recruiter-in-the-loop semantic matching, full override discretion, audit-friendly data export. Sign up free.