Hot topic in Albany: How should NYS regulate AI without stifling innovation?

ALBANY — In the list of buzzwords this legislative session, artificial intelligence might be second only to affordability.

Fresh off the heels of the Responsible AI Safety and Education Act, which lays out rules for the most advanced artificial intelligence models, the state is pursuing a bevy of new laws and regulations.

Legislators say they are seeking to balance the need for regulation with encouraging investment and innovation in emerging technology. Meanwhile, Gov. Kathy Hochul has her own proposals, including legislation aimed at protecting children and further investment in a statewide consortium of research institutions.

In 2025, the state passed a slew of regulations related to artificial intelligence, including legislation outlawing child sexual abuse images child pornography generated with artificial intelligence and artificial intelligence-generated explicit images or “deepfakes” of a person without their consent.

And last month Hochul agreed to sign an amended version of the RAISE Act. In announcing the agreement, Hochul said the legislation would set a “strong and sensible standard,” for cutting-edge AI models and hold developers accountable.

During her State of the State address Tuesday, Hochul said she would support new legislation to ban political campaigns from spreading AI-generated images of people, including opposing candidates, without their consent in the 90 days before an election. She also said she would require clear labels on all AI-generated content.

Called "provenance data," this information would allow people to understand more about where content comes from — as well as how real it is.

"People deserve to trust what they see and hear during election season," Hochul said. "So this year we’ll require disclosures on all AI-generated images and videos so New Yorkers know what’s real and what’s not."

State legislators also have ideas on what guardrails the state needs.

Since the end of the last legislative session, nearly a dozen bills have been introduced related to artificial intelligence. One would mandate search engines to disclose when their search results are generated by artificial intelligence. Another would enact a near ban on the use of artificial intelligence in classrooms prior to the ninth grade.

NY A.I. Act

The broadest bill is the NY A.I. Act. The bill would provide guidance for artificial intelligence development, testing and validation. It would also place the burden of responsibility on developers to prove that artificial intelligence products would not cause harm to New Yorkers.

The bill would require businesses, such as banks, that use artificial intelligence in their decision-making to inform customers prior to a “consequential” decision. Customers would be allowed to opt out of the automated decision-making process and have the decision made by a human.

Developers of high-risk AI systems, those that can have a significant impact on a person's well-being, would also need to conduct third-party audits to determine whether the developer has taken reasonable steps to prevent a foreseeable risk of discrimination.

Assemb. Michaelle Solages (D-Elmont), who is sponsoring the bill, said the goal is to install guardrails for consumers and root out potential biases in systems that use artificial intelligence without stifling innovation within the industry.

For example, she said, some companies have started to use artificial intelligence to review resumes for potential hires. And banks have started to use it to vet customers applying for a mortgage.

Solages said her concern is that those systems might disqualify job candidates who have an unexplained gap in their work histories or a potential homeowners whose credit history was damaged by a missed payment years ago.

“There's a difference between a human looking at it and using empathy and using their knowledge, versus an AI algorithm that's determining what's best,” she said.

Difficult balance

The tension between protection and innovation will likely be the biggest struggle as the legislature works through the bills.

Assemb. Ed Ra (R-Franklin Square), said the legislature’s goal should be to encourage new opportunities for business in the state while understanding that artificial intelligence comes with dangers if used improperly.

Ra said he thought the final version of the RAISE Act struck that balance.

Ra said he believed it was important to listen to AI developers as the state crafts additional legislation. That input gives legislators a real understanding of the practical implications new laws might have, he said.

“I always say we are called upon to be experts in everything,” he said. “And if we're honest with ourselves, we know we're not experts in everything.

Ra said he wants to see how artificial intelligence can be harnessed to improve state government, noting that governments often tend to be behind the curve when it comes to the use of technology.

Assemb. Andrew Gounardes (D-Bay Ridge) cautioned against allowing too much industry influence on state legislation.

The RAISE Act is the strongest AI safety law in the country, Gounardes said. But the state has more to do, especially when it comes to “frontier models,” or cutting-edge artificial intelligence systems, he added.

Gounardes said he believes some of the provisions that were cut during negotiations over the bill need to be included in future legislation. Gounardes said he is also focused on artificial intelligence chat bots and consumer safety.

Earlier this month Hochul proposed legislation that would require social media platforms to automatically default to the highest privacy settings for anyone under the age of 18, turning off open chat functions to prevent strangers from messaging minors and disabling location sharing. It also would limit youth interactions with artificial intelligence companions, known as companion bots, which simulate a human relationship with a user.

“I think there's a lot more to be doing in this space," Gounardes said. "I am absolutely, mind-bogglingly terrified about the prospect of AI chatbots in kids’ toys.”

Job killer?

Gounardes said the state needs to start talking about the potential displacement effects of AI on the workforce.

“I think we might be on the cusp of seeing a really huge transformation in terms of job displacement from the evolution of AI, as the technologies get sharper and better, and as they start having more real-world applicability,” he said. “If people can't have a career, people can't have a job to provide for themselves or their families, then we're in a really bad place.”

Ra acknowledged there would be some job displacement as artificial intelligence models improve and become more widespread.

“I think that's inevitable to some extent,” he said. “I think that the flip side is it creates a tremendous amount of opportunity for people who are well-trained in understanding how to use these tools to be part of a new type of job.”

Gounardes said he doesn't believe the state faces a binary choice between regulation and allowing innovation. The state can see the use of artificial intelligence systems that don’t eliminate large sections of the workforce or cause potentially catastrophic harm while putting in regulations that allow the industry to thrive and protect people, he said. 

“Detroit did not put seat belts and air bags in cars out of the goodness of their hearts," he said. "They did it because they were required by regulation to do so.”