NYC Local Law 144: Bias Audits and Notices for AI Hiring Tools

NYC Local Law 144 is New York City's law on automated employment decision tools (AEDTs), the software that scores, ranks or screens job candidates and employees. Since enforcement began on 5 July 2023, employers and employment agencies in the city may only use such a tool for hiring or promotion if it has had an independent bias audit within the past year, a summary of the results is published, and candidates are notified.

The law was the first in the United States to require bias audits of AI hiring tools, and it remains the reference point for HR tech vendors. In December 2025, the New York State Comptroller found that the city's enforcement had been ineffective, and the enforcing agency agreed to tighten its approach. This guide explains what counts as an AEDT, how bias audits and impact ratios work, the notice requirements, penalties, and what employers and vendors should do.

What is Local Law 144?

Local Law 144 of 2021 amended the New York City Administrative Code to regulate the use of automated employment decision tools. It took effect on 1 January 2023, but enforcement was delayed while the Department of Consumer and Worker Protection (DCWP) finalised rules. DCWP adopted final rules in April 2023 and began enforcement on 5 July 2023. The rules define key terms, set out how bias audits must be calculated, and explain notice and publication duties.

What counts as an AEDT?

An automated employment decision tool is any computational process derived from machine learning, statistical modelling, data analytics or artificial intelligence that issues a simplified output, such as a score, classification or recommendation, used to substantially assist or replace discretionary decision making for employment decisions that affect people.

Under the DCWP rules, "substantially assist or replace" means the output is relied on solely, with no other factors considered; is one of a set of criteria where it is weighted more than any other criterion; or is used to overrule conclusions from other factors, including human decision making. Tools that do not meet this test, such as a spam filter or a simple keyword search that does not drive decisions, are generally outside scope. Whether a particular tool is covered is fact-specific and depends on how the employer actually uses its output.

Who must comply?

• Employers using an AEDT to screen candidates for hiring or employees for promotion for positions in New York City.
• Employment agencies, including staffing and recruitment firms, using AEDTs for the same purposes.

The obligations fall on employers and agencies, not directly on vendors. In practice, vendors are asked to commission bias audits of their tools, provide data and support notices, because their clients cannot comply without that help.

Is it mandatory or voluntary?

Mandatory for covered employers and employment agencies. There is no certification. Compliance is shown through a current bias audit, a published summary and documented notices.

Independent bias audit

Before using an AEDT, the employer or agency must ensure it has been the subject of a bias audit conducted no more than one year before use. The audit must be performed by an independent auditor, meaning a person or group that is not involved in using, developing or distributing the tool, has no employment relationship with the employer or vendor, and has no direct financial interest in it.

The audit calculates, at minimum:

• Selection rates for each sex category, each race and ethnicity category, and intersectional categories of sex, race and ethnicity, where the tool selects candidates to move forward.
• Scoring rates where the tool produces scores, meaning the rate at which individuals in a category receive a score above the sample's median.
• Impact ratios: each category's selection or scoring rate divided by the rate of the most selected or highest scoring category.
• The number of individuals assessed who fall into an unknown category.

The rules allow an auditor to exclude categories representing less than 2 percent of the data from impact ratio calculations, with an explanation. Audits should use historical data from the employer's or the vendor's use of the tool; test data may be used if insufficient historical data exists, with an explanation. Multiple employers using the same tool may rely on one audit if it includes their data or historical data from other users.

A worked impact ratio example

Suppose a screening tool reviewed 1,000 applicants and advanced some of them to interview. An illustrative calculation for sex categories might look like this:

Category	Applicants	Advanced	Selection rate	Impact ratio
Male	520	156	30.0%	1.00
Female	460	115	25.0%	0.83
Unknown	20	5	Reported as a count	Not calculated

The most selected category has an impact ratio of 1.00, and others are compared with it. The same calculation is repeated for race and ethnicity categories and for intersectional categories such as Hispanic or Latino female. Local Law 144 does not set a pass or fail threshold, but many employers use the four-fifths (0.80) guideline from US federal selection procedures as a warning sign that warrants investigation. The numbers above are illustrative only.

Publishing the results

Before using the tool, the employer or agency must make publicly available on its website a summary of the most recent bias audit, including the date of the audit, the source and an explanation of the data used, the number of individuals in each category, selection or scoring rates and impact ratios, and the distribution date of the tool. The summary must be clearly identified and easy to find, and it must stay posted for at least six months after the tool's last use.

Notices to candidates and employees

Employers and agencies must notify candidates who reside in New York City and employees being considered for promotion:

• That an AEDT will be used in assessing them, and that they may request an alternative selection process or a reasonable accommodation under other laws, if available.
• The job qualifications and characteristics the tool will use in the assessment.
• Information about the type of data collected, its source and the employer's data retention policy, if not disclosed on the website, available on written request within 30 days.

Notice must be given at least 10 business days before use, and can be provided on the careers website, in the job posting, or by email or mail.

Penalties and enforcement

Violations carry civil penalties of up to $500 for a first violation and each additional violation on the same day, and between $500 and $1,500 for each subsequent violation. Each day an AEDT is used in violation, and each failure to give notice to a candidate, can be a separate violation. DCWP enforces the law, mainly in response to complaints. The law does not prevent individuals from bringing claims under other anti-discrimination laws.

The December 2025 State Comptroller audit

On 2 December 2025, the New York State Comptroller published an audit of DCWP's enforcement of Local Law 144 covering July 2023 to June 2025. It concluded that enforcement had been ineffective. According to published summaries, test calls to the 311 hotline about AEDTs were frequently misrouted, DCWP's review of 32 published bias audits found one issue while the Comptroller found at least 17 potential issues, and DCWP had not followed its own enforcement procedures or consulted the city's technology office. DCWP agreed to most recommendations, including better complaint routing, staff training, written review procedures, and use of interviews and tool demonstrations in enforcement. Employers should expect more scrutiny of published audits and notices.

How employers comply, step by step

1. Inventory tools used in hiring and promotion, including ATS features, assessments, video interview scoring and chatbots that screen candidates.
2. Determine coverage by examining how each tool's output is used in decisions.
3. Obtain a bias audit from an independent auditor, using the vendor's or your own historical data.
4. Publish the summary on your careers site before use, and keep it current.
5. Set up notices at least 10 business days before use, with a way to request an alternative process or accommodation.
6. Handle data requests within 30 days.
7. Repeat audits annually while the tool is in use.
8. Review results and consider whether adverse impact needs action under anti-discrimination law.

What vendors should provide

• A current independent bias audit of the tool, with data from multiple customers where possible.
• Exports of selection or scoring outcomes by demographic category for customer-specific audits.
• Clear documentation of what the tool does, its inputs, outputs and intended use.
• Configuration options and text for candidate notices and alternative process requests.
• Information on data collected, sources and retention for customer disclosures.
• Version tracking so audits match the tool version in use.

Documentation you will need

• AEDT inventory and coverage analysis.
• Bias audit reports and auditor independence confirmation.
• Published summary and posting history.
• Candidate notice templates and delivery records.
• Records of alternative process and data requests.
• Vendor contracts and documentation.

Common mistakes

• Assuming a tool is out of scope because a human makes the final decision, when the human mostly follows the tool's ranking.
• Using a vendor audit without checking it covers the version and use case.
• Summaries buried or missing required elements.
• Late notices, given at the time of assessment rather than 10 business days before.
• Audits older than one year.
• Ignoring demographic data quality, with large unknown categories.

Best practices, tools and technologies

• Collect voluntary demographic data separately from decision data, with clear consent.
• Log every automated output with tool version, for audits and investigations.
• Monitor impact ratios continuously, not only annually.
• Use the four-fifths rule from US employment guidance as an internal warning signal.
• Give recruiters guidance on how much weight to give tool outputs.
• Structure AI governance with the NIST AI RMF or ISO/IEC 42001.

We build recruitment platforms with outcome logging, audit data exports, notice workflows and alternative process requests through our job board and recruitment and custom software work.

Other US rules on AI in hiring

New York City is not alone. Illinois has required notice and consent for AI analysis of video interviews since 2020 and amended its Human Rights Act, effective 1 January 2026, to address discriminatory use of AI in employment decisions and require notice. California's Civil Rights Council adopted regulations on automated decision systems in employment that took effect in October 2025, and California's privacy regulator has rules on automated decision-making technology. Colorado's automated decision law takes effect in 2027. Federal anti-discrimination laws apply throughout. Employers hiring across states should design one approach to notices, audits and human review that meets the strictest applicable rule.

How it relates to other laws

Law or framework	Relationship
Colorado AI Act	Broader state law on automated decisions, including employment, effective 2027.
EU AI Act	Classifies AI used in recruitment and promotion as high-risk, with detailed obligations.
CCPA/CPRA	California rules on automated decision-making technology, including in employment.
Federal anti-discrimination law	Title VII and related laws apply to disparate impact from any selection procedure, automated or not.
NIST AI RMF	Framework for managing fairness and other AI risks.

Real-world example

A New York staffing agency uses a resume-ranking tool built by an HR tech startup with an Indian engineering team. The startup commissions an independent audit using anonymised historical data from several clients, publishes the results for clients to reference, and adds an export of selection outcomes by category. The agency confirms recruiters advance candidates mainly based on the ranking, treats the tool as an AEDT, posts the audit summary on its careers page, adds a notice to job postings 10 business days before screening begins, and offers a link to request an alternative process.

Next steps

1. Inventory hiring and promotion tools.
2. Decide which are AEDTs.
3. Get or refresh independent bias audits.
4. Publish summaries and set up notices.
5. Prepare for closer enforcement.

Frequently asked questions

What is NYC Local Law 144?

A New York City law requiring bias audits, public summaries and candidate notices for automated employment decision tools used in hiring and promotion.

When did enforcement start?

On 5 July 2023, after the law took effect on 1 January 2023 and DCWP adopted final rules.

Does the law apply to vendors?

The obligations fall on employers and employment agencies, but vendors usually provide audits and data to support them.

How often is a bias audit required?

The tool must have had an audit within one year before each use, so in practice annually.

Who can perform the bias audit?

An independent auditor not involved in using, developing or distributing the tool and without a financial interest in it.

What is an impact ratio?

A category's selection or scoring rate divided by the rate of the most selected or highest scoring category.

How much notice must candidates get?

At least 10 business days before the tool is used.

What are the penalties?

Up to $500 for a first violation and $500 to $1,500 for subsequent violations, with each day of use a separate violation.

Does the law require changing a biased tool?

It requires the audit and disclosure; it does not itself require specific actions based on results, but anti-discrimination laws still apply.

Does a human in the loop remove a tool from scope?

Not if the tool's output is relied on solely, weighted more than other criteria, or used to overrule other factors.

What did the 2025 state audit find?

That DCWP's enforcement was ineffective; DCWP agreed to improve complaint handling, training and review procedures.

Do remote candidates outside New York City need notice?

The notice duty applies to candidates who reside in New York City; employers often give the same notice to everyone for simplicity.

Can several employers share one bias audit?

Yes, if the audit includes their data or uses historical data from other employers using the same tool, as the rules allow.