Understanding NYC Local Law 144: Regulating Automated Employment Decision Tools

As artificial intelligence (AI) and automation continue to shape recruitment and the workplace, worries about bias, transparency, and accountability are on the rise. In response, New York City took a pioneering step by enacting a law to regulate Automated Employment Decision Tools (AEDTs), known as NYC Local Law 144 of 2021 or NYC Bias Audit Law. This law requires that, beginning July 5, 2023, all employers and employment agencies utilizing these tools for hiring or promotions must comply with certain auditing and notification protocols.
What is LL 144?
NYC bias audit law, formally the Automated Employment Decision Tool Law, is the initial American law requiring bias audits for AI-based hiring technologies. It impacts employers and employment agencies in New York City using automated tools to assess, screen, or rank job seekers for hiring or promotion.
The objective is to make sure that these tools do not inadvertently perpetuate racial, sexual, or other forms of bias based on characteristics that are protected by relevant law.
Scope of the Law
Who Is Required to Comply?
- Employers and employment agencies within New York City.
- Applies whether or not the company’s headquarters is within NYC, as long as the hiring/promotion decision impacts a position within NYC.
- Utilization of AEDTs must be in the service of a hiring or promotion decision.
What Tools Are Included?
An Automated Employment Decision Tool (AEDT) within NYC Bias Audit Law is defined as:
“Any computational operation. that produces simplified output, such as a score, classification, or recommendation, used to significantly aid or replace discretionary decision making.”
This encompasses software that pre-screens resumes, scores job candidates, or suggests applicants as long as they are applied to significantly aid or replace decisions made by humans.
Key Requirements of NYC Bias Audit Law
Annual Bias Audits
- Employers need to make sure that any AEDT they use has had an independent bias audit within 1 year of using this legitimate tool.
- The audit must also investigate differential effects—within sex, race, and/or ethnicity as an example of categories that may apply more generally to many uses —using historical data or test data.
- The audit must be conducted by an accredited, third-party auditor (who was not involved with the development or use of AEDT).
- Metrics Required:
Selection rate: The proportion of candidates selected from within each demographic group.
Impact ratio = Selection rate comparison – if rates are different, the same group does worse across measures.
Public Disclosure
- Employers must investigate and publicly disclose the Corporate Bias Audit on their website.
- They must also disclose:
(a)The date of the audit
(b) Sex, race and ethnic distribution
(c) Data used in the audit- origin and level of data
Candidate Notification
- Candidates must be notified at least 10 business days in advance that:
(a)An AEDT will be used.