In the absence of meaningful action on the federal level, state lawmakers are moving New Jersey to be one of the first states to ensure artificial intelligence companies can be held legally – and financially — liable for any harms caused by their technologies.
A group of Democrats in the state Assembly have proposed legislation that would modernize state liability law to specifically cover crimes committed by AI products or agents.
The AI liability bill was introduced earlier this year, but several recent high-profile incidents involving hacking and other alleged crimes by autonomous AI agents have elevated its importance in the eyes of many groups pushing for reform.
“You can’t chase technology, but you can establish fundamental principles of regulation, which happen to be best done through product liability,” said Beverly Brown Ruggia, who serves as the financial justice program director at New Jersey Citizen Action and is also a Consumer Federation of America board member.
Earlier this week, the Assembly’s Science, Innovation and Technology Committee took up the issue of AI regulation during a lengthy hearing held inside the State House in Trenton. While no votes were cast during the hearing, it featured testimony from a host of national and state experts.
Many suggested the federal government’s general lack of action in this area means state policymakers must step in to address what some are calling an “accountability crisis.”
“We know what fraud looks like, we know what theft looks like, we know what murder looks like,” Ruggia said during the hearing. “If that happens, somebody is responsible, and somebody needs to be held accountable. It’s no different (with) AI.”
A comprehensive report (opens in new tab) published in 2024 by a special task force convened by former Gov. Phil Murphy — who was a strong backer of artificial intelligence — highlighted the potential for numerous societal benefits from the use of AI technology. Among them was the ability to put advanced capabilities into the hands of many users lacking technical training.
However, there’s also been a strong public pushback across the state in more recent months, including in response to the construction of energy-consuming data centers that are needed to support many AI products.
A poll conducted earlier this year by Stockton University’s William J. Hughes Center for Public Policy revealed a majority of the registered voters who were surveyed oppose the construction of data centers where they live.
Meanwhile, a plurality of 41% said they believe society’s increased use of AI will make their lives worse, according to the poll, which had a margin of error of +/- 3.9%.
New Jersey policymakers, in recent years, have moved to enact several laws governing AI technology. These include the criminalization of the use of realistic AI-generated images and videos known as deepfakes in incidents involving harassment, extortion or some other unlawful purpose.
Among the additional reforms now up for consideration in the state Assembly is the AI liability measure, which sponsors are calling an “accountability act” for generative artificial intelligence technology.
If enacted, the owners, operators and developers of AI platforms could be held financially liable if their products or agents are, among other things, “engaging in conduct which, if it had been committed by a human,” would constitute the commission of a crime, according to the bill (opens in new tab).
The civil penalties that could be assessed under the bill would total as much as $5 million per violation, on top of compensatory and punitive damages, according to one of the proposed amendments provided to NJ Spotlight News.
To be sure, experts say current product liability law already likely applies to the products being developed by AI companies.
But state Assemblyman Andrew Macurdy (D-Union) — the bill’s primary sponsor — said cases involving such claims against AI companies can take a long time to work their way through the courts. The proposed legislation would establish what’s known as strict liability, which is an enhanced legal standard, Macurdy said.
“This bill really is the next level, beyond what’s already on the books,” he said during a recent interview with NJ Spotlight News.
Speaking more broadly, Macurdy said state lawmakers’ efforts to create sensible guardrails for emerging AI technologies is part of taking on what he considers to be “the public policy question of our time.”
“With a federal government that has not taken any real steps toward that, it’s up to the states to put good ideas forward, and to push the conversation,” Macurdy said.
“When we have people at the top of the industry indicating there should be regulation, I think that we owe it to our current generation, and to future generations, that we listen, and step up, and try to put some comprehensive ideas forward,” he went on to say.
During the hearing, Macurdy listed a number of high-profile incidents that have already occurred involving generative AI.
These included what’s commonly referred to as the Hugging Face incident, which gets its name from a company that was hacked over the summer by autonomous AI agents developed by another company, OpenAI. The agents were able to bypass safeguards that were supposed to prevent them from getting online, according to published reports.
“It doesn’t take much imagination to envision increasingly serious harms as the technology evolves in the months and years ahead,” Macurdy said during the hearing. “That is, unless AI companies are given reason to proceed more carefully.
“If the companies face the real possibility of civil liability that impacts their bottom line, they will take stronger measures to ensure their products do not hack other companies or otherwise harm the public.”
Approaching AI regulation through product liability law is “the single best way to change (the) behavior of AI companies,” said Rob Eleveld, cofounder of the Seattle-based Transparency Coalition, a group that advocates across the nation for AI policies that promote safety, transparency and the public good.
New Jersey is one of a small number of states currently pursuing the adoption of legislation seeking to make it clear that the liability laws apply directly to AI products, he said during an interview with NJ Spotlight News.
This means New Jersey could be the first state in the nation to enact such legislation, depending on how quickly lawmakers here and in other states can get their respective legislation over the finish line.
“This is an important bill, and it will put New Jersey at the head of the list, with a group of other states, I think,” Eleveld said during the interview.
He predicted as many as 10 states could take action in this area over the next year or so.
“I think every state Legislature should be moving with urgency right now because every day that goes by, the AI companies are pouring, literally, billions of dollars into this,” Eleveld went on to say.
The proposed legislation in New Jersey, though, is facing opposition from one of the state’s leading business-lobbying groups.
Among the concerns raised by the New Jersey Business & Industry Association representative Jack Ramirez during this week’s hearing was how the proposed penalties could impact smaller and emerging AI companies, including those operating in New Jersey.
Ramirez credited bill sponsors for including what’s known as an affirmative defense clause in the legislation, something the bill says could protect companies that are making “reasonable efforts to prevent the GAI platform from violating the provisions of the bill.”
But Ramirez urged lawmakers to provide a fuller definition of what those “reasonable efforts” could be.
“We look forward to working with your team in order to hopefully get this into a much more business friendly light,” Ramirez told Macurdy during the hearing.
Ramirez also raised concerns about other bills included in the Assembly’s package of AI legislation.
These included a measure seeking to force AI companies to create and place labels on “any image, video, or audio content . . . created or altered by the covered provider’s generative artificial intelligence system.”
Another piece of legislation would require some AI developers to prepare and submit certain disclosures to the state Attorney General’s Office, including annual risk management protocols.
None of the individual bills were put up for votes in committee this week, but state Assemblyman Chris Tully, who chairs the Science, Innovation and Technology Committee, suggested it’s only a matter of time before at least some of the measures are adopted.
“This is a topic that takes time, that is of consequence,” Tully (D-Bergen) said during the final moments of this week’s hearing.
“I think what the public is looking for is regulation, and they’re looking for their legislators and their policymakers to act, and this committee will certainly act,” Tully said.