Bias testing under NYC LL 144
A bank that uses an automated tool to hire for New York City roles may take on obligations under Local Law 144, depending on how the tool is used. This page sets out what the law asks of the bank and how Suited's testing maps to it.
Local Law 144 of 2021 governs the use of automated employment decision tools in hiring and promotion in New York City. The Department of Consumer and Worker Protection has enforced it since July 5, 2023, and its page on the law links the rules and an FAQ.
The four requirements
- A bias auditWithin one year before the tool is used, by an independent auditor who was not involved in using, developing or distributing it and has no employment relationship or financial interest that the rules exclude1
- Impact ratiosSelection rates or scoring rates across sex categories, race and ethnicity categories and their intersections2
- A published summaryBefore use, the audit's date, its data source and explanation, the number of people assessed in each category, including an unknown category, the rates and impact ratios the audit calculated and the distribution date, on the employment section of the website for at least 6 months after the tool's latest use. The rules add disclosures for audits that used test data or left a very small category out3
- NoticeAt least 10 business days before use, to candidates and employees who live in New York City: on the careers page, in the job posting or by mail or email for candidates, and in a written policy or procedure, the job posting or by mail or email for employees considered for promotion4
The notice also explains which qualifications and characteristics the tool assesses and how to request an alternative selection process or a reasonable accommodation under other laws, where one is available. The bank also posts the type and source of the data and its retention policy on the employment section of its website, with instructions for requesting them in writing, and answers a written request within 30 days.
Whether it applies to your bank
The law covers a tool that substantially assists or replaces discretionary decision making. The rules define that as relying on the tool's output alone, weighting it more than any other criterion or using it to overrule conclusions drawn from other factors, including human decision making. Whether the law applies depends on how the bank uses the tool in its own process, so that analysis sits with the bank and its counsel.
How Suited's testing maps to it
Suited tests every bank-specific model for adverse impact across race and gender before it is deployed, using 4/5ths-rule selection-rate ratios from the EEOC Uniform Guidelines, backed by standardized effect sizes. Models that produce unjustified disparate impact are not deployed.
Every model also has an independent bias audit before deployment and annually after that, whether or not the bank's use of Suited falls under Local Law 144. Suited's models are audited by Holistic AI and have been found to show no evidence of disparate impact, on a standalone or an intersectional basis.
Demographic data is optional for candidates and never used in scoring. Suited collects it so that every model can be tested.
What the bank owns
The obligations under the law sit with the employer. Suited supplies scores and the audit work behind them. It does not accept or reject candidates, and bankers make every hiring decision. The bank decides how scores enter its process, publishes the audit summary where the law requires one and gives candidates the required notice.
Before each recruiting cycle, confirm four things with the people who own them: counsel's view of how the program uses each tool, the date of the most recent audit, the published summary on the careers site and the notice that goes out 10 business days before use. Ask your Suited team for the most recent audit summary for your bank's model when you start.
See the data on your bank.
The walkthrough covers what the retention data shows about your bank and the talent landscape around it. Sixty minutes, built to brief a division head. Live, not a deck.
Book a data walkthrough