New York Launches State-Run Institute to Test AI Models, Opens Registration Under New Safety Law

New York on Tuesday became the first state to open a government-run testing program for advanced artificial intelligence systems, as Governor Kathy Hochul launched a research institute at Binghamton University and the state's financial regulator switched on an online portal requiring large AI developers to register ahead of a new safety law taking effect January 1, 2027.

The announcement, made at a press event in Albany, combines two pieces of New York's response to the rapid commercial deployment of AI: a new academic center that will independently evaluate frontier AI models, and the first operational step in enforcing the Responsible AI Safety and Education Act, known as the RAISE Act, which Hochul signed into law last December.

What was announced

The new institute, called the New York Center for AI Responsibility and Research, will be based at Binghamton University, part of the State University of New York system, and run by interim co-directors Greg Morrisett, dean of Cornell Tech, and Jeannette Wing, executive vice president for research at Columbia University. According to Binghamton University's own description of the center, it is intended to make the campus "a leading force in open, accountable, public-focused AI research," drawing on more than 80 researchers from institutions across the state.

The center's funding traces to a $30 million philanthropic gift announced in January, led by Binghamton alumnus and Bloomberg LP co-founder Tom Secunda, who chairs the state's Empire AI consortium, combined with a $25 million research investment from SUNY and additional operating capital from the state. Computing capacity will come from Empire AI, the $500 million public-private partnership New York built to give public researchers access to large-scale AI infrastructure.

At the same event, the state's Department of Financial Services, acting through a newly created Office of Digital Innovation, Governance, Integrity and Trust, known as DIGIT, opened the registration portal required under the RAISE Act. The law applies to "large frontier developers," defined as AI companies with more than $500 million in annual revenue — a threshold that, as TechCrunch reported, squarely covers OpenAI and Anthropic. Those companies, and others like them, must register with DIGIT before January 1, 2027, after which they must publish safety frameworks on their own websites, issue public model transparency reports and notify the state of critical safety incidents within 72 hours.

A law with teeth, and a narrow one

The RAISE Act was signed in December 2025 after New York legislators softened an earlier draft to resemble California's AI safety statute, in part to reduce the odds of a successful federal preemption challenge. It does not regulate unreleased or experimental models, and it stops short of requiring companies to hand over proprietary model weights or training data to the state. Its leverage instead comes from disclosure and penalties: companies that fail to register or that violate reporting requirements face fines of up to $1 million per violation, rising to $3 million for repeat violations, though a state spokesperson declined to specify how enforcement actions would be brought.

That structure explains why Hochul paired the regulatory portal with a research institute. The RAISE Act can compel companies to publish safety information, but DIGIT, a small new office inside DFS with only one dedicated full-time hire so far, has limited capacity to independently verify what companies disclose. The Center for AI Responsibility and Research is designed to fill that gap, examining published safety frameworks and transparency reports and feeding findings back to regulators and other public officials.

"The stakes are too high to simply take AI companies at their word," Governor Hochul said in announcing the center, adding at the event that New York would use the registration system to gain "more visibility into the inner workings of these AI companies."

DFS Acting Superintendent Kaitlin Asrow, whose agency houses DIGIT, said in a statement that "DFS is proud to lead the implementation of the RAISE Act" through the new office. Empire AI chair Tom Secunda said independent testing "will help researchers understand what increasingly powerful systems can do and how to use them safely," while SUNY Chancellor John B. King Jr. and Binghamton University President Anne D'Alleva both framed the center as a platform for the university system to work on AI safety at scale.

Who is affected, and who is not yet weighing in

The immediate compliance burden falls on a short list of companies large enough to clear the $500 million revenue threshold, chiefly OpenAI and Anthropic among AI-specific developers, though the rule is written broadly enough to reach any sufficiently large frontier-model developer operating in New York. Smaller AI startups, academic labs and companies that only deploy third-party models without training them are not covered. Neither OpenAI nor Anthropic had issued public statements responding to the registration requirement as of Tuesday evening, and TechCrunch's report on the law's rollout did not include comment from either company or from AI safety advocacy groups.

New York is not alone in moving on frontier-model oversight. California's SB 53, which took effect in January, already requires large AI developers to maintain safety frameworks, protect whistleblowers and report critical incidents within 15 days — a faster clock than New York's 72-hour window but arguably a lower bar given the longer reporting period applies there to a broader set of incidents. Illinois has a similar law, SB 315, that takes effect at the start of 2027 and goes further by requiring independent safety compliance audits, a feature New York's law does not mandate but which the new Binghamton center could effectively provide on a voluntary or state-directed basis.

What happens next

Large frontier developers now have until January 1, 2027, to complete registration with DIGIT, after which the agency is expected to begin reviewing the safety frameworks and transparency reports companies are required to publish. DIGIT, which sits inside DFS and was still building out its staff as of its most recent public update, is expected to add additional full-time employees "in the coming weeks and months," according to the agency's own earlier announcement of the registration timeline. The Center for AI Responsibility and Research is conducting a national search for additional faculty and permanent leadership beyond its two interim co-directors, and its first public findings have not yet been scheduled.

The arrangement is also a test of how far a single state can push AI oversight absent federal rules. New York's law was signed in defiance of a Trump administration executive order aimed at limiting state-level AI regulation, and legal challenges on preemption grounds remain widely anticipated among technology and telecommunications lawyers who have been tracking the statute since its passage. Whether the RAISE Act survives that test, and whether the new Binghamton center produces findings that alter how regulators or the public view frontier AI systems, will likely shape whether other states beyond California and Illinois adopt similar structures in 2027.