AI Compliance · 2026-07-30 · 10 min read
NYC Local Law 144 Bias Audit: Employer Guide for 2026
NYC Local Law 144 fines don't stay at $500 — audit and posting failures accrue per day, and notice failures accrue per candidate. Here's who the law covers, what counts as an AEDT under the "substantially assist or replace" test, and the four duties every employer owes.
A NYC Local Law 144 bias audit is an independent, annual evaluation of any automated employment decision tool (AEDT) used to screen candidates or employees for jobs in New York City. NYC Local Law 144 has been in force since January 1, 2023, with enforcement since July 5, 2023. The employer — not the software vendor — must commission the audit, publish a results summary, and notify candidates before use.
New York City moved before every US state on AI hiring, and its penalty mechanics are unforgiving: audit and posting failures accrue per day, notice failures per candidate. Here is who is in scope, what counts as an AEDT, the four duties, the math, and how to comply.
Does NYC Local Law 144 apply to my company?
Local Law 144 applies to employers and employment agencies that use an AEDT to screen candidates for hire or employees for promotion for jobs in New York City. There is no size threshold: a five-person startup ranking resumes with an AI tool for one Manhattan role is covered.
The NYC nexus test
The law follows the job, not your headquarters. You are in scope when the role being screened is tied to New York City — in practice, when the position is performed at least part of the time from an office in the city, or is a fully remote role associated with a New York City office. The candidate-notice duty runs to candidates and employees who reside in the city.
- A Texas-headquartered company hiring for a hybrid role in its Manhattan office is covered for that role.
- A Manhattan-headquartered company hiring for a role based entirely in its Austin office generally is not covered for that role.
If one ATS screens every requisition company-wide, check where each screened role sits — headquarters geography is irrelevant.
What counts as an AEDT? The "substantially assist or replace" test
The statute defines an AEDT as a computational process — derived from machine learning, statistical modeling, data analytics, or artificial intelligence — that issues a simplified output (a score, classification, or recommendation) used to substantially assist or replace discretionary decision-making in hiring or promotion.
The city's enforcement rules narrowed that middle phrase considerably. Under the Department of Consumer and Worker Protection (DCWP) implementing rules, a tool "substantially assists or replaces" discretionary decision-making only when its output is:
- Relied on exclusively, with no other factors considered;
- Weighted more heavily than any other criterion; or
- Used to overrule conclusions drawn from other factors, including human judgment.
That narrowing is why many employers concluded their tools fall outside the law — but what counts is how a tool is actually used, not what the policy manual says. Pull your last 50 hires: if no one ever advanced against the tool's recommendation, the tool is deciding, and the exclusive-reliance prong is in play.
| Tool and usage | Likely covered? | Why |
|---|---|---|
| Resume ranker whose score determines who gets an interview | Yes | Relied on exclusively for the screen-out decision |
| Video-interview scoring used as the primary gate | Yes | Weighted above any other criterion |
| AI matching treated as one input among several, with real human review | Often no | Output is not dominant and does not overrule human judgment |
| Keyword search, spreadsheet filters, scheduling automation | No | No simplified predictive output driving the decision |
The four employer duties under Local Law 144
| # | Duty | Core requirement | Penalty accrual |
|---|---|---|---|
| 1 | Independent bias audit | Less than one year old at the time of each use | Per day |
| 2 | Public results summary | Posted on the employment section of your website | Per day |
| 3 | Candidate notice | At least 10 business days before AEDT use | Per candidate |
| 4 | Data disclosure | Data type, source, and retention policy — posted or on request | Tied to the notice and disclosure provisions |
1. Annual independent bias audit (less than one year old at use)
The audit must be performed by an independent auditor — no financial interest in the employer or vendor, no role in developing or using the tool. It calculates selection rates and impact ratios for the tool's outputs across sex and race/ethnicity categories, including intersectional categories.
"Annual" undersells the actual rule: the audit must be less than one year old at the time of each use — a rolling deadline. An audit dated March 10 is stale on March 10 of the following year, and the per-day clock starts.
2. Public summary of results
Before using the AEDT, post a summary of the most recent audit results — including the audit date — on the employment section of your website, along with the tool's distribution date, and keep it up while the tool is in use. An audit that exists but was never posted still generates daily posting violations.
3. 10-business-day candidate notice
Candidates and employees who reside in New York City must be told, at least 10 business days before the AEDT is used on them, that an automated tool will assess them and which job qualifications and characteristics it evaluates, with instructions for requesting an alternative selection process or accommodation. The notice must go out before the tool runs — a trap for teams whose ATS scores every application instantly on submission.
4. Data disclosure
Make information available about the data the tool collects — its type, source, and your retention policy — either posted publicly or provided on request. It is the easiest duty to forget because it lives in privacy documentation, not the hiring workflow.
How do Local Law 144 penalties add up?
The sticker price looks mild: $500 for a first violation, $500–$1,500 for each subsequent violation. The accrual mechanics are what make it expensive — per day for audit and posting failures, per candidate for notice failures.
| Scenario | Violations | Maximum exposure |
|---|---|---|
| AEDT used for 90 days after the audit expired | 90 daily violations | $500 + (89 × $1,500) = up to $134,000 |
| Results summary never posted, tool in use 180 days | 180 daily violations | $500 + (179 × $1,500) = up to $269,000 |
| 500 NYC candidates screened with no 10-day notice | 500 per-candidate violations | $500 + (499 × $1,500) = up to $749,000 |
The streams stack, too — an expired audit, a missing summary, and skipped notices accrue three independent violation streams at once.
Your vendor's ATS is still your problem
Local Law 144 regulates employers and employment agencies, not software vendors. The employer owns the duty even when the AEDT is a feature inside a vendor's ATS or assessment platform. "Our vendor said it's compliant" is not a defense.
- Get the vendor's bias audit and check the date — it must be under a year old at your time of use, by a genuinely independent auditor.
- Confirm the audit covers your configuration, not just the default model.
- Put LL144 duties in the contract — refreshed audits annually, plus notification when a model change invalidates the current one.
- Never assume the vendor sends the notices. The 10-business-day notice and the data disclosure are workflow obligations on your side.
A structured vendor review catches most of this before procurement — our AI vendor risk assessment questions include the hiring-tool items for every questionnaire.
How to comply with Local Law 144: 7 practical steps
- Inventory every tool that touches hiring and promotion decisions, including features buried inside your ATS and tools recruiters adopted on their own — a shadow AI discovery process surfaces the ones nobody registered.
- Apply the AEDT test to each tool as actually used, documenting which of the three DCWP prongs, if any, the usage hits.
- Map the NYC nexus — flag every requisition performed in or tied to a New York City office, including associated remote roles.
- Commission or obtain the independent bias audit, verifying auditor independence and coverage of the required categories and impact ratios.
- Publish the results summary on your employment site with the audit date and tool distribution date.
- Automate the 10-business-day notice and data disclosure in job postings or the application flow, so no NYC-resident candidate is scored before the clock runs.
- Calendar the rolling refresh at month 10, not month 12 — the audit must be under a year old at every use, and auditor lead times are real.
How does Local Law 144 compare to other AI hiring laws?
NYC was first, but no longer alone — the same AEDT can trigger several regimes at once:
| Law | Status | What it adds beyond LL144 |
|---|---|---|
| NYC Local Law 144 | In force since Jan 1, 2023 (enforcement since Jul 5, 2023) | Bias audit + posting + notice + data disclosure |
| Illinois HB 3773 (775 ILCS 5/2-102(L)) | In force since Jan 1, 2026 | Civil-rights exposure via IDHR charges; zip-code proxies expressly prohibited; employee notice duty |
| California FEHA ADS regulations (2 CCR §§11008.1, 11009(f)) | In force since Oct 1, 2025 | Applies at 5+ employees with one CA employee or applicant; anti-bias testing evidence weighable; 4-year records (§11013(c)) |
| Connecticut PA 26-15 | Anti-bias evidentiary rule from Oct 1, 2026; AEDT deployer notices from Oct 1, 2027 | Anti-bias testing becomes evidence in discrimination claims |
| Colorado SB 26-189 | Applies from Jan 1, 2027 | Pre-use disclosure, adverse-decision explanation, correction and human-review rights, 3-year records |
For the full picture, see our complete map of US state AI laws in 2026 and the California AI laws 2026 guide. Operating in Europe too? EU AI Act Article 50 transparency duties apply from August 2, 2026 — check your exposure in minutes with our free EU AI Act checker.
Shieldra tracks NYC Local Law 144 alongside eight other AI-specific frameworks with deterministic, citation-backed classification, so one assessment shows every regime your hiring stack triggers. See what's included on the features page.
FAQ
How often do we need a new bias audit under Local Law 144?
The audit must be less than one year old at the time of each use of the AEDT — a rolling deadline, not an annual calendar task. If it expires while the tool keeps screening, violations accrue per day at $500 for the first and $500–$1,500 for each subsequent day. Most compliant teams re-engage their auditor around month 10.
Does Local Law 144 apply to fully remote jobs?
It can. The law covers jobs located in New York City, which in practice includes fully remote roles associated with a New York City office, and the candidate-notice duty covers candidates who reside in the city. A remote role tied to your NYC office is in scope even if the hire never visits.
Can we rely on our vendor's bias audit?
Often yes, as a practical matter — but the legal duty stays with you. Confirm the audit is under a year old at your time of use, the auditor is genuinely independent, and the audit covers the tool as you configured it. The public posting, the 10-business-day notice, and the data disclosure remain your obligations regardless.
What happens if the bias audit shows disparate impact?
Local Law 144 does not prohibit using a tool with unfavorable impact ratios — it requires the audit, the summary, the notice, and the data disclosure. But publishing poor ratios creates ready evidence for discrimination claims under laws like Illinois HB 3773 (775 ILCS 5/2-102(L)). Most employers treat a bad audit as a signal to reconfigure or replace the tool.
Who enforces NYC Local Law 144 and what are the penalties?
The NYC Department of Consumer and Worker Protection enforces the law, and has since July 5, 2023. Penalties are $500 for a first violation and $500–$1,500 for each subsequent one — accruing per day for audit and posting failures, and per candidate for notice failures. Because the duties accrue independently, exposure for one neglected tool can reach six figures within months.