Gov. Kathy Hochul wants lawmakers to scale back a bill that could restrict the release of AI models, according to three people familiar with negotiations, in a shift that would bring New York’s rules more in line with a new California law.
Both the California law and New York’s Responsible AI Safety and Education Act, or RAISE Act, require developers to disclose safety information and steps they’ve taken to prevent severe risks, like an AI model that is used to develop biological weapons or commit massive fraud.
But the bill on Hochul’s desk would prevent developers from deploying new models “if doing so would create an unreasonable risk of critical harm.”
Lobbyists for AI companies and the tech sector have been making their case to Hochul, a Democrat who has championed state investments in AI computing, to drop that language because they say it’s vague and could stifle innovation. The governor is also facing pressure from parents and consumer advocates who say the current AI landscape is like the wild west and she should hold firm.
President Donald Trump complicated the deliberations when he signed an executive order last week pre-empting states’ ability to regulate AI. The Republican president said a patchwork of regulations could threaten the country’s ability to be a leader in the emerging technology; he wants to let Congress act and ensure a clear path forward for companies.
The people familiar with Hochul’s proposed changes said they aren’t authorized to discuss them publicly because the closed-door negotiations are ongoing.
A spokesperson for Hochul said in a statement that she wants the state’s regulatory approach to “be a model for the nation.” Last week, the governor signed a bill that would require ad-makers to disclose if their commercials include AI-generated synthetic performers.
Hochul said last week that she doesn’t want Washington limiting the state’s ability to protect residents, but acknowledged the executive order is now part of her decision-making.
“We are an AI supportive state,” Hochul told reporters on Dec. 8. “We have a lot of ideas on how we want to manage this. It would be devastating if the federal government removed our ability as leaders of our states to protect our citizens. It should come from them. They should be focusing on the same thing we are.”
Advocates of the RAISE Act call it common-sense regulation. The bill passed over the objections of some Republicans. It would require companies to develop safety plans on how they will prevent models from being used to develop chemical, biological or nuclear weapons and ensure models don’t operate without “meaningful human intervention.”
The bill would require developers to report violations to the state and would let the attorney general sue developers for violations of the law by their models. The attorney general could seek an injunction preventing a model’s release.
“If their own testing shows that one of their models poses an unreasonable risk, they shouldn't release their model,” said Assemblymember Alex Bores, a Democrat from the Upper East Side who sponsored the bill. “That's based on avoiding a situation that we had with the tobacco companies, where they were the first to know that cigarettes cause cancer, but they kept denying it publicly and releasing their products.”
Bores is part of a crowded field to succeed U.S. Rep. Jerry Nadler in Congress next year. Bores said his advocacy for AI regulation has drawn the interest of some deep-pocketed people in the tech industry. They’ve pledged $100 million to fund a super PAC called Leading the Future.
The group released an ad last week that said adopting the RAISE Act “creates a chaotic patchwork of state rules that would crush innovation.” Leading the Future spokesperson Josh Vlasto said the group was looking for a consistent national regulatory framework.
Hochul can take a step toward doing that by mirroring California, which is implementing SB53. That law requires developers to disclose an AI framework that includes a discussion of how a model would prevent catastrophic risks. The California legislation applies to a smaller set of developers and doesn’t include any language that would restrain the release of new models.
Julie Samuels, president of the advocacy group Tech:NYC, said the California law is workable and should be copied in New York.
“There would be a de facto national standard,” she said. “Every major tech company is in New York and California. So, all of a sudden, now you see responsible Blue State governors taking the reins and that's great.”
Hochul’s spokespeople wouldn’t elaborate on her exact negotiations, but three people familiar with the talks said she submitted almost the exact text of SB53.
Bores said California’s law was a “good first step that states should build upon.” He said conversations with Hochul’s office are ongoing.
A group of parents is pushing the governor to hold firm, including Mary Rodee, who has seen how hard it is to regulate technology.
Since her teenage son died by suicide in 2021 after an online extortion scam, she’s traveled to the U.S. Capitol and written letters to social media executives. Rodee joined dozens of parents organized by the Tech Oversight Project in pleading with Hochul to sign the legislation with minimal changes.
“Why do they get to fight safety protocols?” Rodee said in an interview. “If you can't do it on the street, you shouldn't be able to do it in these apps. And I just, I really, really hope that the parents and the people can be more powerful and have a stronger voice than these tech-industry groups.”
Hochul has until Friday to act.
Her spokesperson, Kassandra White, said, “Governor Hochul has been at the forefront of the innovation economy and remains committed to advancing AI responsibly as she reviews the legislation.”
