NYC Local Law 144 and AI note-takers: what is in scope

New York City's Local Law 144 bars employers and employment agencies from using an automated employment decision tool (AEDT) to screen candidates or employees for employment decisions in the city unless the tool has had an independent bias audit within the past year, a summary of the results is public, and candidates get notice at least ten business days before it is used. For interview note-takers, one detail in the city's rules matters most: output from tools that "transcribe a video or audio interview" is not the kind of output the law targets, while tools that score, rank or recommend candidates may be, depending on how you use them.

This guide sets out the definitions carefully, applies them feature by feature to AI interview tools, and lists practical steps. It does not decide for you whether your tool is covered; that depends on facts only you have.

What Local Law 144 requires

The law took effect on January 1, 2023, and DCWP began enforcement on July 5, 2023. Primary sources: the NYC Administrative Code, §§ 20-870 to 20-874, DCWP's final rules (6 RCNY Subchapter T), the DCWP AEDT page and the DCWP FAQ.

What counts as an automated employment decision tool

The statute defines an AEDT as "any computational process, derived from machine learning, statistical modeling, data analytics, or artificial intelligence, that issues simplified output, including a score, classification, or recommendation, that is used to substantially assist or replace discretionary decision making for making employment decisions that impact natural persons" (§ 20-870). Break that into four tests, all of which must be met:

1. The technique. DCWP's rules say this means mathematical, computer-based techniques that generate a prediction or a classification, and "for which a computer at least in part identifies the inputs, the relative importance placed on those inputs, and, if applicable, other parameters" (6 RCNY § 5-300).
2. Simplified output. A prediction or classification, which may take the form of a score, tag, recommendation or ranking. The rules add that it "does not refer to the output from analytical tools that translate or transcribe existing text, e.g., convert a resume from a PDF or transcribe a video or audio interview" (§ 5-300).
3. Substantially assists or replaces discretionary decision making. Defined narrowly; see the next section.
4. For employment decisions. The statute means decisions "to screen candidates for employment or employees for promotion within the city," and the rules define "screen" as determining whether someone "should be selected or advanced in the hiring or promotion process."

The statute also excludes tools that do not automate, support, substantially assist or replace discretionary decision-making and do not materially impact people, giving examples such as a junk email filter, spreadsheet or database. DCWP's FAQ adds that using a tool to scan a resume bank, conduct outreach or invite applications does not trigger the law, because that is not screening candidates who have applied.

"Substantially assist or replace": the three ways a score counts

Producing a score is not enough on its own. Under 6 RCNY § 5-300, a simplified output substantially assists or replaces discretionary decision making only if you use it in one of three ways:

This is why the same software can be in scope for one employer and not another. The definition looks at how the output is used in your process, not only at what the software produces. It also means your process can drift into scope: a score that starts as one input among many can, under time pressure, become the thing that decides.

Where AI interview note-takers fall, feature by feature

AI interview tools rarely do only one thing. Apply the definitions to each feature separately. The table below points to the relevant rule text and the question to answer; it does not state a conclusion for your tool.

The same questions apply to any tool, including Interview Signal, whose scorecards attach ratings to evidence quotes from the transcript. Under the rules, what matters is how your team uses those ratings in the decision.

Two cautions. First, the transcription carve-out sits in the definition of simplified output; it does not say a product that also transcribes is outside the law. Second, the law applies to "employers" and "employment agencies" that use an AEDT, so an agency using a scoring feature for a client's NYC role has its own obligations, whatever the vendor says.

Who and where: jobs, candidates and agencies

DCWP's FAQ says the law applies to use of an AEDT "in the city," which it describes as any of these:

• The job location is an office in NYC, at least part time.
• The job is fully remote but the location associated with it is an office in NYC.
• The employment agency using the tool is located in NYC, or, if it is outside NYC, one of the two conditions above applies.

The notice obligation runs to candidates and employees "who reside in the city." A "candidate for employment" is someone who has applied for a specific position by submitting the necessary information in the required format, so sourcing and outreach sit outside it.

The bias audit, in practice

If a tool is in scope, the audit requirements in 6 RCNY §§ 5-300 to 5-303 include:

• Independence. The auditor cannot have been involved in using, developing or distributing the tool, cannot be employed by the employer or vendor, and cannot have a direct or material indirect financial interest in either.
• Calculations. Selection rates (or scoring rates, for tools that score) and impact ratios for sex, race/ethnicity and intersectional categories, plus the number of people in an unknown category. A category representing less than 2% of the data may be excluded, with a justification.
• Data. Historical data from use of the tool; test data only if there is not enough historical data. DCWP's FAQ says imputed or inferred demographic data cannot be used.
• Timing. You may not use the tool if more than one year has passed since the most recent audit.
• Publication. A summary with the data source and explanation, the number of people assessed in the unknown category, the number of applicants or candidates, the selection or scoring rates and impact ratios for all categories, the audit date and the tool's distribution date.

A vendor's audit can use historical data from several employers, but under § 5-302(a) you may rely on an audit built on other employers' data only if you gave the auditor your own historical data from using the tool, or you have never used the tool. Ask a vendor for its audit summary, its date and whose data it used, and check it against that rule.

Notices: timing, delivery and content

For candidates, 6 RCNY § 5-304 allows notice by any of three routes, at least 10 business days before use:

• the careers or jobs section of your website,
• the job posting, or
• U.S. mail or email.

The notice must include:

• that an AEDT will be used in assessing the candidate,
• the job qualifications and characteristics the tool will use, and
• instructions for requesting an alternative selection process or a reasonable accommodation under other laws, if available.

The rules add that "nothing in this subchapter requires an employer or employment agency to provide an alternative selection process." The statute also requires information about the type and source of data the tool collects and your data retention policy, either on your website or on written request, and § 5-304(d) of the rules sets out how to post that information and request instructions on the employment section of your site and answer requests within 30 days.

An LL144 notice is not a recording consent line, and it has different timing. If you also record or transcribe the interview, you still need the spoken consent step described in the interview recording consent script.

Enforcement since 2023

On December 2, 2025, the New York State Comptroller released an audit of DCWP's enforcement of Local Law 144 (Report 2024-N-6, covering July 2023 to June 2025). It reported that DCWP received two AEDT complaints, found its complaint routing ineffective, and noted that DCWP identified a single issue of non-compliance across 32 companies reviewed while the auditors identified at least 17 instances of potential non-compliance. DCWP generally agreed with the recommendations while disputing parts of the draft.

As of September 2026 we found no amendment to §§ 20-870 to 20-874 in the Administrative Code, and DCWP's FAQ is still dated June 2023. Separately, Local Law 25 of 2026 requires a study and report on the impacts of algorithmic tools and AEDTs on employees; it does not change Local Law 144's requirements. Thin enforcement so far is not a reason to skip compliance: penalties accrue per day, and the law preserves candidates' own right to sue.

Practical steps for teams hiring in New York City

1. Inventory. List every tool that touches candidates for NYC-linked roles, and every feature in each, using the feature table above.
2. Write down how outputs are used. For each score or recommendation: is it ever the only factor, the most heavily weighted factor, or used to overrule people? Put the answer in your hiring procedure, not just in someone's head.
3. Ask vendors specific questions. Does the feature produce a score, classification, ranking or recommendation? What identifies the inputs and their weight? Is there a bias audit, when, by whom, on whose data? Can scoring be switched off while keeping transcription?
4. Decide with counsel, feature by feature, and record the reasoning and date.
5. If in scope, arrange or obtain the audit, publish the summary, and put the notice in postings or on the careers site at least 10 business days before use.
6. If out of scope, keep it that way: train interviewers that tool scores are one input and never override the panel, and review the process yearly.
7. Check other rules too. Hiring in Illinois brings its own AI notice rules (see the Illinois AI Video Interview Act guide), and recording consent still applies everywhere (see is it legal to record a job interview).