New York Launches First State-Run AI Testing Center to Independently Evaluate Frontier Models

New York has officially launched the nation's first state-run artificial intelligence testing program, shifting oversight from corporate self-reporting to independent scientific evaluation. Governor Kathy Hochul inaugurated the New York Center for AI Responsibility and Research at SUNY Binghamton on November 10, 2026. The initiative aims to verify the safety claims of frontier AI developers.[1][4]

The center brings together a statewide network of more than 80 researchers from institutions across New York. Their mandate covers assessing how advanced systems operate, testing their security, and developing measurement tools for increasingly powerful models. The collaborative effort is designed to give public officials objective evidence for regulatory decisions.[3][4]

"While the federal government fails to act, New York is leading the way on AI safety and holding big tech accountable," Governor Hochul stated during the launch. "The stakes are too high to simply take AI companies at their word." She emphasized that the state is leveraging its academic strength to protect residents.[3]

Greg Morrisett, dean and vice provost of Cornell Tech, and Jeannette Wing, executive vice president for research at Columbia University, will serve as the center's interim co-directors. They are tasked with guiding the scientific agenda and overseeing the independent evaluations. The researchers will remain at their home institutions while participating in the statewide network.[1][3]

The center's research agenda extends beyond immediate safety testing to include foundational scientific inquiries. Teams will investigate formal methods, hardware security, and the interpretability of complex neural networks. By understanding the mathematical theory behind these models, researchers hope to predict and mitigate unintended behaviors before they manifest in real-world applications.[1][3]

AI

Harnessing Supercomputing for Public Good

The testing center is powered by Empire AI, a $500 million public-private supercomputing initiative designed to support research for the public good. This infrastructure provides the massive computational capacity required to run comprehensive safety evaluations on large-scale frontier models. The combination of academic independence and state-of-the-art hardware positions New York uniquely in the regulatory landscape.[3][4]

Tom Secunda, chair of the Empire AI board and co-founder of Bloomberg LP, highlighted the strategic importance of the computing resources. "We are harnessing the power of Empire AI and the collaborative strength of our academic institutions to advance AI research," Secunda said. He noted that independent testing will clarify what powerful systems can actually do.[3]

The center's foundation was laid with a $30 million philanthropic gift from a donor group led by Secunda, alongside a $25 million research capital investment from the State University of New York. The state government is also providing operating capital to sustain the facility. The November launch expands that initial vision into a fully operational network.[4]

Researchers at the center will tackle complex technical challenges, such as preventing AI agents from escaping theoretical testing sandboxes. By developing novel technical innovations, the facility aims to establish rigorous standards for model interpretability and alignment. The work represents a shift toward using science to understand how models can be utilized safely.[1]

The RAISE Act Compliance Timeline

The launch of the testing center coincides with the opening of a state registration portal for the Responsible AI Safety and Education Act. The legislation, which takes full effect on January 1, 2027, establishes strict transparency and reporting requirements for developers of frontier models. The law targets systems trained using massive computational power.[4][5]

Under the RAISE Act, large frontier developers generating more than $500 million in annual revenue must register with the state before the January deadline. The newly formed Office of Digital Innovation, Governance, Integrity, and Trust, housed within the Department of Financial Services, will oversee compliance. The office has already published guidelines for developers operating within New York.[1][4]

Beginning in January 2027, large developers will be required to publish comprehensive safety frameworks on their public websites. These documents must detail how the companies identify, assess, and mitigate catastrophic risks associated with their models. The mandate forces developers to formalize and disclose the internal protocols they use to manage advanced AI systems.[1][4]

All frontier developers operating in the state will also face strict incident reporting obligations. The law requires companies to report critical safety incidents to the state office within 72 hours of discovery. Additionally, developers must publish model transparency reports and submit quarterly assessments of catastrophic risks to state regulators.[1][4][5]

Expanding Regulatory Authority

While the initial RAISE Act sets a baseline for transparency, New York lawmakers are already proposing a significant expansion of regulatory authority. Legislators introduced the RAISE Act 2.0 in late October 2026, granting state officials the power to restrict computing resources utilized by major AI developers. The proposal reflects growing concerns over the rapid pace of model development.[2]

The revised legislation would require covered developers to submit quarterly reports detailing how much of their AI research is automated. This includes disclosing whether existing models are being used to accelerate the development of future, more powerful systems. The expanded definition aims to capture well-funded companies that have spent over $1 billion on development.[2]

State Senator Andrew Gounardes, a sponsor of the new legislation, argued that recent incidents demonstrate the necessity of stronger government oversight. "It's time to rein in the AI arms race," Gounardes said in a statement. "Every day, frightening new headlines make it clear: when it comes to AI, we can't trust Big Tech companies to regulate themselves."[2]

The proposed expansion would also mandate independent audits of frontier models, aligning with the state's broader push for third-party verification. By potentially limiting computing power, regulators hope to establish a mechanism to pause or slow development if catastrophic risks are identified. The legislative push underscores New York's proactive stance in managing technological disruption.[2]

A National Model Amid Federal Inaction

New York's regulatory framework emerges as the federal government steps back from mandatory AI safety rules. The Trump administration recently opted for a voluntary framework, prompting states to fill the governance void. New York joins California and Illinois in enacting some of the nation's first laws governing frontier artificial intelligence models.[3][5]

A recent report from the American Enterprise Institute analyzed the converging approaches of these three states, noting that they all reject licensing and pre-deployment approval in favor of mandated transparency. The report highlights that mandated transparency holds developers accountable to their own safety commitments. It provides governments with actionable evidence without stifling innovation.[5]

The American Enterprise Institute report points out meaningful differences between the state laws. While California gives developers 15 days to report a critical safety incident, New York and Illinois enforce a much tighter 72-hour window. Furthermore, Illinois is set to become the first state to require annual independent third-party compliance audits starting in 2028.[5]

The report also notes that the compute and revenue thresholds used by these states are weakening proxies for actual risk. As models become more efficient, highly capable systems can be trained using fewer resources, potentially falling below the regulatory triggers. Policy experts suggest that future amendments may need to focus on demonstrated capabilities rather than just training compute.[5]

However, the state-level approach does present limitations. The Binghamton testing center will primarily rely on the information that AI companies are legally required to disclose or choose to share publicly. Confidential models that are still in the early stages of development will largely remain inaccessible to the independent researchers until they approach deployment.[1]

Despite these constraints, the combination of the RAISE Act's reporting mandates and the testing center's technical capacity creates a robust oversight ecosystem. By requiring rapid incident reporting and funding independent scientific evaluation, New York is establishing a precedent for AI governance. The state's model balances the drive for technological advancement with the imperative of public accountability.[4][5]

The success of the initiative will depend on the center's ability to recruit top-tier talent and develop effective evaluation methodologies. As the January 2027 compliance deadline approaches, the AI industry will be watching closely to see how the state enforces the new regulations. The outcome will likely influence how other jurisdictions approach the challenge of governing frontier technologies.[3][4]