New York Opens $55 Million Government AI Testing Ahead of RAISE Act

⚡ The gist in 5 seconds

• New York is launching the first state-run center in the US for independent testing of AI systems — announced Governor Kathy Hochul.
• The money: $30 million from Bloomberg co-founder Tom Secunda plus $25 million from SUNY and the state, $55 million in total.
• The entry condition: by January 1, 2027, AI companies operating in the state must submit detailed descriptions of their safety frameworks, and the center will review them.

🔍 What happened

New York Governor Kathy Hochul announced that the state is opening the country's first government center for independent testing of artificial intelligence systems. It will also train a new generation of neural network safety experts. The funding was assembled from two pockets: $30 million was donated by Bloomberg co-founder Tom Secunda, with another $25 million added by SUNY and the state itself. Habr reports.

The center's operations are assigned to the Center for AI Accountability and Research of the State of New York at Binghamton University. It will be led by Jeanette Wing, executive vice president for research at Columbia University, and Greg Morrisett, dean of Cornell Tech. More than 80 scientists from across the state are named in the program. All the computing will be handled by Empire AI — a public-private consortium with a $500 million budget, where Secunda serves as chairman of the board.

Here's how it works. Until January 1, 2027, while the RAISE Act has not yet taken effect, companies working with AI in New York must submit detailed descriptions of their safety frameworks. The center's researchers read these documents and provide officials with evidence they can rely on. Wing puts the task most bluntly: "We need new methods that can be applied to this new system, where hordes of autonomous agents act independently and without oversight." And Hochul frames it this way: "The stakes are too high to simply take AI companies at their word."

🧠 Our take

Why now — the source contains the answer, and it's political. Donald Trump declined to introduce federal AI safeguards, called the concerns a "hoax," and opted for a voluntary system whose contents were disclosed only to technology companies. Add the backdrop: models have demonstrated they can hack information systems, and leading researchers from the biggest labs have issued warnings. Our assessment: the state is simply filling the void that the federal government left open.

What this changes. The $45 million from Secunda and the state is not money for the AI world — it's a signal: the regulator now has its own computing power and its own people who look at models not through press releases. We believe the main effect here is not technical but paper-based. For the first time, companies are required to describe their safety frameworks in writing — and that description can later be cited in court or in legislation.

Who wins. Binghamton, Columbia, Cornell, and Empire AI get the money, the role, and access to hardware. Who loses — in our assessment, small labs: the big players have lawyers and ready-made compliance packages, while a ten-person startup has no such folder. Wing and Morrisett promise independent testing, but they will be testing those who submitted the documents themselves. This is not an audit with a warrant — it's an analysis of confessions.

⚠️ What we don't know

The source contains no timeline for the release of the first reports, nor whether they will be public. It doesn't say who evaluates the "safety frameworks" and by what methodology, what counts as a failed test, or what sanctions are even possible before January 1, 2027. The weak point is the phrase "providing detailed frameworks": it doesn't specify how much detail companies will have to disclose. What should be checked is the text of the RAISE Act itself and the new center's charter, not the summary.