The NYC Local Law 144 bias audit: what it must contain
On this page
Under NYC Local Law 144, an employer or employment agency may not use an automated employment decision tool on a candidate or employee for a job in New York City unless the tool has had an independent bias audit within the past year, a summary of that audit is published on the employer's website, and required notices have gone out to candidates. DCWP has enforced the law since July 5, 2023. This page is about what the audit itself must contain and publish; for what counts as an AEDT in the first place, including where interview note-takers and transcription tools fall, see NYC Local Law 144 and AI note-takers.
This is not legal advice. It summarizes Local Law 144 of 2021 and DCWP's final rule, codified at Title 6, Chapter 5, Subchapter T of the Rules of the City of New York, as published on nyc.gov, as of September 2026. Confirm your specific audit and publication obligations with counsel or a qualified auditor.
What the audit has to calculate
The bias audit's core calculation is the impact ratio, run for each protected category the U.S. Equal Employment Opportunity Commission's EEO Component 1 report uses:
- Calculate the selection rate for each sex category, race/ethnicity category, and each intersectional category (for example, Black male, Asian female) among candidates who applied for the job. For tools used to score, rather than select, candidates the tool must instead calculate a scoring rate, defined as the rate at which candidates in a category receive a score above the sample's median score.
- Identify the category with the highest selection or scoring rate.
- Divide every other category's rate by that highest rate to produce the impact ratio for each category.
An impact ratio of 1.0 means a category was selected at the same rate as the top-selected category; a ratio below 0.80 mirrors the EEOC's four-fifths guideline for adverse impact, though Local Law 144 itself does not set a compliance threshold; a low ratio is a number the audit reports, not a violation by itself. See Title VII disparate impact and AI hiring tools for how the four-fifths test is used in a discrimination claim generally.
The rule permits calculating the audit using either the employer's own historical data or data supplied by the AEDT's developer from other clients' use of the tool, and it allows excluding a category from the calculation if the sample size is not statistically significant, provided the summary discloses that exclusion and its reason rather than omitting the category silently.
Who counts as an independent auditor
DCWP's rule defines independence by conflict, not by credential. An auditor must not:
- Have been involved in using, developing or distributing the AEDT being audited.
- Have any employment relationship with the employer or employment agency, or with the AEDT's developer, beyond performing the audit.
- Have any financial interest in the employer, the employment agency, or the AEDT's developer.
The rule does not create a licensing regime or a government-approved list of auditors; the employer is responsible for confirming and being able to show that whoever performed the audit meets these independence requirements. A vendor's in-house "compliance team" auditing its own AEDT does not meet this bar; a third-party firm with no stake in the tool or its results does.
What must be published, and where
Before using an AEDT, the employer or employment agency must make the following available on the employment section of its public website:
- The date of the most recent bias audit.
- A summary of the results, including the selection or scoring rates and the impact ratios for each category the audit calculated.
- The source and explanation of the data used for the audit, including the number of applicants and whether historical or test data was used, and if any category was excluded for sample size, that fact.
- The distribution date of the AEDT, if the employer did not conduct the audit itself, so a reader can match the published summary to the correct version of the tool.
The summary must stay posted, and DCWP's guidance treats the ordinary practice as keeping it up for as long as the tool remains in use, refreshed with each new annual audit; check the current FAQ on nyc.gov for the specific duration DCWP expects at the time you publish, since this is the kind of operational detail regulators update more often than the underlying rule text.
A 2025 review by the New York State Comptroller's office found gaps in what companies had actually posted, citing missing impact ratios, incomplete demographic breakdowns, and summaries omitting required information at a number of the companies it reviewed. Posting something under a "Local Law 144" heading is not the same as posting a summary that contains every element the rule requires; check your own published page against the four bullet points above line by line, not against whether a page exists at all.
The candidate and employee notice, separately from the audit
Section 20-871(b) requires a separate notice, at least 10 business days before using an AEDT, to a candidate or employee residing in New York City who applies for a position or promotion covered by the tool. Under DCWP's rule, an employer can satisfy this by posting notice on the careers section of its website, including it in a job posting, or providing it directly (such as by mail or email) within that 10-business-day window. The notice must say that an AEDT will be used, describe the job qualifications and characteristics it will assess, and give instructions for requesting an alternative selection process or a reasonable accommodation under other law, where one is available. The law does not itself require an employer to offer an alternative process; it requires telling candidates how to ask, and any duty to actually provide one comes from other law, such as disability accommodation requirements.
Enforcement since 2023
DCWP began enforcing Local Law 144 on July 5, 2023. Civil penalties run $500 for a first violation and each violation occurring the same day as the first, and $500 to $1,500 for each subsequent violation. Each day an employer uses a non-compliant AEDT is treated as a separate violation, and a missing or defective candidate notice is a separate violation from a missing or defective bias audit summary, so exposure compounds quickly on a tool used continuously without a current audit. This is a civil penalty structure, not a criminal one, and DCWP, not a private plaintiff, brings the enforcement action.
A checklist to run before your next audit is due
- Confirm the date of your last bias audit for each AEDT used on NYC candidates or employees, and calendar the next one at least a month before the one-year mark.
- Confirm your auditor has no financial interest in, and was not involved in developing, the tool being audited.
- Pull up your own published summary and check it against all four required elements: audit date, selection or scoring rates and impact ratios by category, data source and sample explanation, and distribution date.
- If any demographic category was excluded for sample size, confirm the summary says so and why, rather than simply omitting it.
- Confirm your candidate notice goes out at least 10 business days before the tool is used, and that it names the qualifications assessed and how to request an alternative process or accommodation.
- Keep a log of when each notice went out and to whom, separately from the audit paperwork, since DCWP treats them as separate compliance obligations with separate penalties.
Common mistakes
| What people assume | What the rule requires |
|---|---|
| "A bias audit is just a vendor's fairness statement" | It must be conducted by someone independent of both the employer and the tool's developer, with no financial interest in either. |
| "Publishing once is enough" | A new audit, and a refreshed published summary, is required at least annually while the tool remains in use. |
| "An impact ratio under 0.80 is automatically illegal" | Local Law 144 requires calculating and publishing the ratio; it does not itself set a pass/fail line the way an enforcement action under Title VII might use the same number differently. |
| "The notice and the audit are the same requirement" | They are separate duties under separate provisions, enforced as separate violations with separate penalties. |
| "Having a page that mentions Local Law 144 is compliant" | The city's own 2025 review found many published summaries missing required elements despite a page existing. |
Questions people ask
What is an impact ratio in a Local Law 144 bias audit?
It is the selection rate (or scoring rate) for a demographic category divided by the selection rate of the highest-scoring category in the same audit, calculated separately by sex, race/ethnicity, and the intersection of both. An impact ratio below 0.80 mirrors the EEOC's four-fifths rule and signals potential adverse impact, though the law itself does not set a pass or fail line.
Who can perform an independent bias audit?
DCWP's rule requires the auditor to be independent of the employer and of the AEDT's developer: someone with no financial interest in the AEDT or the outcome of the audit, who was not involved in developing or deploying the tool being audited, and who has the relevant expertise to conduct the audit.
Does the bias audit need to be redone every year?
A new bias audit is required at least annually, within one year of the last one, for as long as the employer keeps using the AEDT. Using a tool with an audit older than a year is itself a basis for a violation, independent of what the audit found.
Can a small candidate pool exempt an employer from publishing certain results?
The rule allows an employer to exclude a category from its published impact ratios if the sample for that category is not statistically significant, but the summary must still disclose which categories were excluded and why, so an employer cannot silently drop unfavorable numbers.