On behalf of our member companies that make NJBIA the largest, most impactful state-level business association, we write to express our concerns with S-4075 and respectfully seek amendments to address significant compliance and operational issues raised by the bill.
We understand and support the intent behind S-4075: ensuring that emerging technologies do not become a vehicle for unlawful workplace discrimination. But discrimination is already illegal in New Jersey, regardless of whether a decision is made by a person or an algorithm. The New Jersey Law Against Discrimination (NJLAD) prohibits both intentional discrimination and employment practices that have an unlawful disparate impact on protected classes. In fact, the Division on Civil Rights has specifically clarified that these protections apply to artificial intelligence and automated decision-making tools.
If there are gaps in those protections created by new technology, we welcome a conversation about how to close them. S-4075, however, goes far beyond closing a discrimination loophole.
The bill creates an expansive new regulatory structure governing employers’ use of technology, including independent impact assessments, public reporting, employee notice, ongoing monitoring, data-retention requirements, human review, appeal procedures and new avenues for liability. The cumulative effect is not simply to prohibit discriminatory uses of AI, but to regulate how employers use a broad range of technology in routine workplace decisions.
The breadth of the bill’s definitions makes that especially concerning. An “automated employment decision system” is defined broadly enough to include a “computational process” that uses employee information together with machine learning, statistical modeling, data analytics, artificial intelligence or similar methods to produce outputs used in employment-related decisions. And those
employment decisions include not merely hiring or termination, but decisions affecting work schedules, compensation and work tasks.
As written, those definitions could capture routine workplace technologies that employers already use every day. A spreadsheet or software program that analyzes staffing, payroll or scheduling data could potentially qualify simply because it uses “data analytics” to inform a workplace decision. Nor does the technology need to make the decision itself—the bill can apply when its output is merely used by an employer in making that decision.
This means that even routine uses of workplace technology could subject an employer to a complex regulatory regime. Employers could be required to commission independent impact assessments, satisfy public reporting and employee notice requirements, establish ongoing monitoring and human-review procedures, and create mechanisms for employees to challenge technology-assisted decisions. These are significant compliance obligations, particularly for businesses without dedicated legal, compliance and technology teams.
The burden is compounded by requirements that may depend on information employers do not possess. Businesses purchasing widely available, off-the-shelf technology may have little ability to obtain proprietary information about how those products were developed, trained or tested, yet the employer may still bear responsibility for complying with requirements tied to that information.
More fundamentally, S-4075 moves toward a system of government pre-approved technology use. Rather than identifying specific harmful or discriminatory uses and prohibiting them, Section 3 limits covered technology to specified purposes and further requires its use to be “necessary” to accomplish those purposes. Employers must also navigate standards involving concepts such as employee “dignity,” “autonomy” and “welfare.” The question for a business therefore becomes not simply whether its use of technology is lawful or nondiscriminatory, but whether the State has authorized the purpose for which it is being used and whether the employer can demonstrate that its use was sufficiently “necessary.”
That approach creates substantial uncertainty for businesses deciding whether to adopt new technologies. Terms like “dignity,” “autonomy,” “welfare” and “necessary” do not establish clear compliance lines, particularly when applied across thousands of different workplaces and technologies. Faced with that uncertainty, employers may simply decline to use beneficial technologies rather than risk discovering after the fact that a regulator viewed their use differently.
That uncertainty is especially concerning given the consequences of noncompliance. S-4075 does not rely solely on civil enforcement. Section 18 provides that a willful violation may constitute a criminal offense carrying a mandatory minimum $1,000 fine and the possibility of up to 100 days of imprisonment. Criminal penalties are difficult to reconcile with a regulatory framework that simultaneously relies on broad definitions and subjective standards.
New Jersey does not need to choose between protecting workers from discrimination and allowing businesses to responsibly adopt new technology. NJLAD already prohibits discriminatory employment decisions, including practices producing unlawful disparate impacts, and the State has expressly recognized that those protections extend to automated tools.
Rather than building an expansive regulatory system around virtually every workplace use of data analytics or AI, we respectfully encourage the Legislature to identify specific harms that existing law does not address and target those gaps directly. Doing so would protect employees while providing employers the clarity necessary to comply, innovate and continue adopting technologies that are increasingly part of everyday business operations.
Thank you for considering NJBIA’s perspective on this bill. Please email either Jack Kelly at jkelly@njbia.org or Jack Ramirez at jramirez@njbia.org if you would like to further discuss our position.
