New York Establishes Nation's First Independent AI Safety Testing Center With 80 Researchers

New York AI Takes Lead on Independent Safety Testing

Governor Kathy Hochul launched the nation's first state-run AI testing program on Tuesday, establishing the New York Center for AI Responsibility and Research at the State University of New York (SUNY), Binghamton University1

1

3

3

. The initiative marks a departure from federal reliance on voluntary safeguards, with Hochul stating, "The stakes are too high to simply take AI companies at their word"1

1

.

More than 80 New York scientists will contribute to independent AI safety testing through the center, using computing power from Empire AI, the state's $500 million public-private AI partnership1

1

. Cornell Tech Dean Greg Morrisett and Columbia University Executive Vice President for Research Jeannette Wing will serve as interim co-directors3

3

.

Funding and Research Infrastructure Drive AI Regulation

The center's funding includes a $30 million philanthropic gift from donors led by Bloomberg Co-Founder and Binghamton alumnus Tom Secunda, an initial $25 million SUNY research investment, and state operating support3

3

. Researchers will examine whether AI systems are safe and secure, can be manipulated, can produce dangerous results, or can compromise privacy2

2

.

Participating researchers will continue working at their home institutions while developing tools, measurements, and standards for evaluating increasingly capable systems3

3

. The research agenda targets experts focused on foundational limits and mathematical theory, formal methods and verification, agentic and complex systems, hardware security and privacy, interpretability and alignment, and scientific AI and biosecurity3

3

.

RAISE Act Implementation Creates Accountability Framework

The center's work connects directly with implementation of the RAISE Act, which takes effect January 1, 20273

3

. New York's RAISE Act, signed by Kathy Hochul last year, requires large AI developers to register with the state by January 1, 20271

1

. Starting then, companies must publish safety frameworks and transparency reports and report critical safety incidents within 72 hours1

1

.

Hochul's office announced an online registration portal administered by the Office of Digital Innovation, Governance, Integrity, and Trust, known as DIGIT3

3

. At the state's request, the center can provide expertise, assessments, and evidence to inform AI oversight3

3

.

Addressing Emerging Threats and Complex AI Systems

The announcement came one week after midterm elections revealed public anxiety about AI governance. About six in 10 voters said the technology would do more to hurt society, according to the AP Voter Poll, although only about one-quarter called AI a crucial voting factor3

3

.

Morrisett highlighted the urgency: "This center gives us an opportunity to work on really hard problems, like how do we adapt best practices, such as formal methods to prevent agents from escaping their sandboxes? Or how do we understand new classes of attacks and failures and the defenses that we need to develop?"2

2

Wing told Politico, "We need new techniques that can be applied to this new kind of system with swarms of autonomous agents acting independently and without control"3

3

.

Limitations and Future Directions for Responsible AI Applications

One potential hole in the program is that it can only study what AI companies disclose publicly1

1

. Researchers will have access to publicly disclosed information, while companies' internal models under development remain private3

3

. The center's remit includes supporting responsible AI applications in medicine, scientific research, and critical infrastructure3

3

.

Hochul's office said it will explore additional ways to strengthen New York's AI safety laws and regulations in the coming months3

3

. The initiative gives regulators an independent basis for evaluating companies' safety claims and identifying shortcomings in their disclosed safeguards, though visibility into every emerging risk remains limited3

3

.