The RAISE Act — shorthand for Responsible AI Safety and Education Act — requires large frontier AI developers operating in New York to create publicly available safety frameworks outlining how they are mitigating risk. They will also be required to submit reports of critical safety incidents that could lead to catastrophic risks, defined as death or serious injury to 50 or more people or at least $1 billion dollars in damages, to the state within 72 hours, as well as quarterly assessments of catastrophic risks. The state attorney general’s office will lead on enforcement of the law.
Developers will be required to pay an assessment for DIGIT’s work in supervising them, similar to how the Department of Financial Services charges banks operating in New York a fee.
In a statement, Gilman said he’s aiming to “create more transparency in this industry and support responsible AI development.”
But even before the RAISE Act is implemented, some are already urging DIGIT to do more to get to that goal.
“That office is empowered to propose additional regulations and/or potentially laws to help improve the frontier of AI safety,” Democratic state Sen. Andrew Gounardes, one of the sponsors of the RAISE Act, said in an interview last week. “I really hope and expect that that office takes that charge seriously and does become an office that has the expertise necessary to provide guidance on what future regulations can and should look like.”
Gounardes, along with state Assemblymember Alex Bores, spearheaded the RAISE Act, which Hochul later negotiated to make it more palatable to the tech industry and more closely aligned to California’s AI legislation.
Bores, who is slated to appear with Hochul on Monday as she rolls out the RAISE Act implementation updates, also said he would like to see DIGIT ask for more data from the tech companies. In addition, he wants to see universal monitoring of all AI that’s deployed, third-party auditors for the companies with employee-level access, a prohibition on developers releasing models that fail their own safety tests and expanded whistleblower protections. Some of those provisions were included in an earlier iteration of the RAISE Act.
“I have always said that we need action at the federal level for AI, but so far the federal government has been completely incapable of doing it and states have been the only level of government that has actually regulated AI,” Bores said. He added that even though he thinks the RAISE Act needs to “go a lot further,” he’s “very pleased to hear the governor talking about the role of states and the role of New York in continuing to push forward AI safety.”