NJ lawmakers weigh new limits on AI use in business decisions
Four bills target artificial intelligence in employment, housing and insurance decisions.
Artificial intelligence is changing how businesses evaluate workers, renters, borrowers, and insurance claims. New Jersey lawmakers are considering four bills that would place new restrictions on how businesses use the technology in employment, housing and insurance decisions.
The Assembly Science, Innovation and Technology Committee advanced three bills Monday addressing insurance claims, tenant screening and mortgage lending. A fourth proposal, which would restrict workplace surveillance and automated employment decisions, was discussed but not voted on. Labor and business representatives offered competing testimony on the bill.
Bill A4981 would restrict the use of AI and other automated systems to make employment decisions, monitor workers, and determine eligibility for public benefits and services.
The bill would not ban workplace surveillance outright. It would prohibit employers from using monitoring tools to interfere with union activity, track employees during off-duty hours, or establish productivity quotas likely to significantly harm worker health and safety. Employers also could not take adverse action against workers based on quotas or performance standards that had not been disclosed.
Employers would be prohibited from requiring workers to install monitoring software on personal devices. They also could not take adverse employment action solely on the basis of data collected through continuous time-tracking tools, such as keystroke loggers, which record typing activity, or idle-time trackers, which monitor when a device is active.
Employment decisions involving these systems would require meaningful human oversight. Workers affected by adverse decisions would have the right to challenge them through a review process. The bill would establish similar protections for people affected by automated decisions involving public benefits and services.
Employers would also have to obtain independent assessments of covered workplace monitoring and decision-making systems before using them and provide advance written notice to workers and unions. In the public sector, agreements to use automated systems or technology likely to result in significant job losses would be subject to collective bargaining requirements.
Union leaders and representatives testified in support of the bill, describing what they said were harmful workplace conditions associated with the growing use of AI.
“Workers are increasingly subject to time-off task metrics that measure every minute of the work day,” said Helen Ireland, the political director for United Food & Commercial Workers Local 360. “Workers will skip bathroom and water breaks to avoid falling behind on AI-generated productivity targets.”
Representatives from industry groups opposed the bill. Some argued that its language was too broad and could unintentionally apply to useful technology unrelated to employee surveillance.
“Hospitals rely on electronic technologies to deliver safe patient care, maintain accurate medical records and protect sensitive patient information,” said Tina Early, the government relations and policy director at the New Jersey Hospital Association. “NJHA is concerned that the legislation’s broader requirements could extend beyond traditional employment-related monitoring and affect systems used for these essential responsibilities.”
Others argued that the required assessments would be too costly and that restrictions on monitoring could discourage employers from allowing staff to work from home.
“If you think about what life is like after COVID, a lot of our sectors, if you’re able to, they all work from home,” said Tony Russo, the president of the Commerce and Industry Association of New Jersey. “This is going to discourage that, because you’re not allowed to use that mouse tracking system or any other kind of system that tracks employees’ whereabouts or positions or what they’re doing.”
Employers convicted of knowingly and willfully violating the law could face fines and jail time. A first offense could carry up to 100 days in jail and fines ranging from $1,000 to $5,000 for each employee harmed. Subsequent offenses could carry 10 to 150 days in jail and fines ranging from $2,000 to $10,000 for each employee harmed.
Employees who successfully sued over violations could recover lost wages and benefits, with interest, plus additional damages of up to 200% of wages owed. Employers and technology vendors could share liability for violations.
The committee heard about 90 minutes of testimony on A4981 but did not vote on the bill.
Three other AI-related bills advanced.
Bill A5494 would prohibit insurance companies from using AI to make the final decision to deny claims under automobile, homeowners, or flood insurance policies. The measure would not prohibit insurers from using AI to assist with other parts of the claims process.
Gary La Spisa, the senior vice president of the Insurance Council of New Jersey, opposed the bill.
“AI is a critical tool for us,” he said. “In the name of keeping down premiums, one of the ways that we can keep down our cost is by improving efficiency. I think there’s a balance to be struck by utilizing and automating and also ensuring we have the right oversight of those tools.”
Assemblymember Andrew Macurdy said he understands that business owners want to use AI to make things more efficient and get costs down.
“I’m on board with that. But like with most things, policy making is a balance,” he said. “Under this bill, you can still use it to do all this massive sorting of the sort of summary denials and tee things up for a human decision maker and gain a ton of efficiencies and cut costs that way. And that to me seems like an appropriate balance.”
Insurers that violated the proposed law could face fines of up to $5,000 per violation.
The two housing bills, A5495 and A5496, would require annual bias audits of automated tenant-screening tools and certain AI systems used to evaluate residential mortgage applications.
Under A5495, landlords would have to notify prospective tenants in writing when automated screening tools would be used to evaluate their applications. The bill would limit screening decisions to measurable factors, such as income and credit scores, and prohibit the use of specified personal characteristics, including race, sex, and gender identity or expression.
Developers of tenant-screening tools would have to implement recommendations from the most recent bias audits and issue public statements explaining how those recommendations were implemented.
Under A5496, mortgage lenders and loan originators would have to notify applicants before they submit applications if generative AI tools would be used to evaluate them. Lenders and loan originators would also have to publish summaries of the most recent bias audits on their websites.
The mortgage bill would require lenders and loan originators to enter measurable information into the tools and prohibit them from entering categories of people into the systems.
Both bills would impose penalties for violations, although the amounts would differ.
Under A5495, developers could face initial civil penalties of up to $1,500, increasing by 2% for each additional day a violation continued. Landlords could face $500 fines for a first offense and fines of up to $1,000 for subsequent offenses.
Under A5496, people who sell, develop, or use covered mortgage tools in violation of specified requirements could face penalties of up to $1,500, with increases for continuing violations. Mortgage lenders and loan originators could face penalties of up to $500 for certain violations, including failures to notify applicants or publish audit summaries.
If enacted, the bills would add to New Jersey’s existing restrictions on AI-related tools in housing. Gov. Mikie Sherrill signed the Forbidding the Algorithmic Inflation of Rent Act into law in July. The law bans algorithmic software that recommends rent prices using competing landlords’ private data.
Committee chair Chris Tully said AI can be a valuable tool when used appropriately, but it is no substitute for human judgement, particularly when decisions can profoundly affect a person’s life.
“As more industries embrace AI, we must establish meaningful oversight and safeguards to ensure these technologies are wielded responsibly, without the risk of error or discrimination,” Tully said. “It is our duty to both guide innovation forward and protect New Jersey residents.”
Curtis Brodner is a Report for America corps member covering housing and affordability for The Jersey Vindicator. He investigates the policies, people, and institutions shaping where New Jersey residents can afford to live, with a focus on accountability and solutions. Previously, he was a criminal justice reporting fellow with Columbia Journalism Investigations, producing investigative work for New York Focus. He earned a master’s degree from Columbia Journalism School as a Toni Stabile Center fellow and a bachelor’s degree in journalism from SUNY Purchase. You can reach him at Curtis AT jerseyvindicator.org.


