NYC Local Law 144 bias audit: a complete guide
- The Local Law 144 bias audit independently tests an AEDT for disparate impact across protected groups.
- It must be conducted by a genuinely independent auditor, and a summary of results must be published.
- The audit needs appropriate outcome data, which organisations must plan to have available.
- It is an ongoing obligation, and a sales-relevant one for vendors who supply hiring AI.
- Current as of June 2026. This is general information, not legal advice.
What the audit tests
Under the DCWP methodology, the bias audit calculates selection-rate or scoring-rate impact ratios by category, comparing outcomes across sex categories, race or ethnicity categories, and intersectional categories, to reveal whether the tool disadvantages particular groups. The analysis is conducted by an independent auditor and is the means by which disparate impact is surfaced.
Who conducts it
The audit must be conducted by an independent auditor, someone not involved in using, developing, or distributing the tool, to ensure objectivity. Independence is central to the requirement's credibility: a self-assessment would not provide the assurance the law seeks. Choosing a genuinely independent, competent auditor is therefore a key step.
The data challenge
A bias audit needs data: information about the tool's outcomes across groups. This raises practical questions about what data is available, how it is collected, and how categories are determined. Employers and vendors often need to plan in advance to ensure the necessary data exists to support a meaningful audit, since you cannot audit for disparate impact without outcome data broken down appropriately.
Publication and notice
The summary of the bias audit results, together with the distribution date of the audit, must be made publicly available on the employer's website, alongside information about the tool. This public transparency is part of the law's design: it lets candidates and the public see that the tool has been audited, when, and what the audit found. Alongside the audit, the law's notice requirements mean candidates must be told when an AEDT is used.
Preparing for and maintaining audits
Because the audit must be done before use and kept current, organisations should treat it as an ongoing obligation, not a one-time event. The tool, the data, and the workforce change over time, so audits must be refreshed. Building the data collection and audit cadence into how the tool is operated is the reliable way to stay compliant rather than scrambling each time.
The vendor dimension
For vendors who supply hiring AI, Local Law 144 is a sales-relevant requirement: employers using your tool in NYC will need to audit it and will look to you for support and data. Being able to help your customers meet the bias-audit requirement, and being transparent about your tool's tested performance, is increasingly part of selling hiring AI. It connects directly to how enterprises assess their AI vendors for governance and risk.
Key terms
- Disparate impact
- A pattern in which selection rates differ across protected groups.
- Impact ratio
- A comparison of selection rates between groups used in bias analysis.
- Independent auditor
- An auditor with no role in using, developing, or distributing the tool.
- Public summary
- The publicly available summary of the bias audit's results.