BILL NUMBER: S4394A
SPONSOR: COMRIE
 
TITLE OF BILL:
An act to amend the labor law, in relation to automated employment deci-
sion tools
 
PURPOSE OR GENERAL IDEA OF BILL:
To limit bias and discrimination against any group on the basis of sex,
race, ethnicity, or other protected class that occurs when automated
employment decision tools are used to screen candidates for hire, and
creating a more specific definition for artificial intelligence auto-
mated employment decision tools in law.
 
SUMMARY OF PROVISIONS:
Section 1- Definitions: Outlines key terminology related to oversight of
artificial intelligence (AI) for employment decision making tools;
Governance: the bill requires deployers or developers to establish,
document, implement and maintain a governance program that contains
reasonable safeguards to map, measure, manage, and govern the foreseea-
ble risk of unlawful discrimination associated with the use of an auto-
mated employment decision tool in addition to staffing requirements;
Transparency: a deployer shall, at all or before the time of a conse-
quential employment decision, notify the individual that is subject to
an employment decision of the tool's use;
Enforcement: the Attorney General may initiate an investigation of a
deployer or developer if a preponderance of the evidence establishes a
suspicion of a violation of these articles;
Compliance: a deployer or developer that complies with these articles
shall be deemed compliant with any law, regulation, rule, requirement,
or standard related to the performance of any impact assessments,
audits, or governance program(s) that are adopted maintained, or
enforced by any political subdivision of New York State.
Section 2 is the effective date.
 
JUSTIFICATION:
In today's job market, employers may receive hundreds of job applica-
tions for a single role. To efficiently evaluate candidates, employers
rely on employment decision tools, like skills tests, resume reading
systems, and automated screening platforms. Many automated employment
decision tools (AEDT) are developed by third-party vendors, yet this
industry of human resources tech providers remains unregulated.
Across the state, employers are not held to any meaningful standard when
disclosing the bias (legally defined as "disparate impact" or "adverse
impact") yielded by these products. As a result, many employers continue
to use hiring tools that disproportionately screen out on the basis of
race and gender and inadvertently bolster resumes and profiles of indi-
viduals with more privileged backgrounds. Additionally, candidates
engaging with these assessments often are not informed about their
purpose or what they are designed to measure.
In 2021 New York City adopted Local Law 144. A Cornell University Report
in late 2024 documented a range of challenges in automating bias audits
under LL 144. The study described difficulties in gathering reliable
demographic data needed for audits, data-quality problems, and issues
ensuring demographic "inclusiveness." Meaning some demographic catego-
ries might not be well represented or correctly reported. There is also
disagreement among auditors, vendors, and companies over what consti-
tutes a valid "independent auditor." The law does not rigidly define who
can audit - which creates a lot of practical uncertainty about audit
quality. Importantly: the law only requires audits for certain
protected categories (race/ethnicity, sex/gender). It does not cover
other axes of bias like age, disability, or socioeconomic background.
Finally, because it's relatively recent, there simply hasn't been enough
time or enough independent empirical research to examine long-term
outcomes: whether historically disadvantaged groups are now getting more
equitable hiring outcomes, or whether disparities have shrunk.
In effect, this bill attempts to do what critics of Local Law 144 say
the City law failed to accomplish: turn bias-audit and transparency into
a more robust, enforceable, systemic framework; with defined obligations
for both employers and vendors, protections for candidates, and a mech-
anism to catch misuse or discriminatory outputs.
This state-level standardization through New York labor laws to govern
use of automated employment decision tools will help clear definitions
and create stronger governance obligations; prohibit a deployer of an
AEDT from using the tool in a way that violates existing anti-discrimi-
nation laws; mandate annual impact assessments and adopt governance
programs to guard against discrimination; notify users of their use and
limitations; prevent employers from relying solely on an AEDT's output
when making hiring, promotion, termination, disciplinary or compensation
decisions; and lastly, provide for enforcement and presumably penalties
or remedies if deployers or developers misuse AEDTs or fail to comply
with requirements.
The aim of this bill is to restore human judgment and labor law compli-
ance in the hiring and promotion process for employers that use AEDTs.
 
LEGISLATIVE HISTORY:
2025: S4394- referred to Labor, amended to A print (restores S5641A
text)
2024: S5641A- referred to Labor
2022: In assembly, advanced to third reading cal.474.
 
FISCAL IMPLICATIONS:
To be determined.
 
EFFECTIVE DATE:
This act shall take effect immediately.