Reported 10 Nov 2026 · Governance A one-minute read
New York will test AI systems itself, weeks before its frontier AI law takes effect.
Governor Kathy Hochul said the state will run independent testing through a university centre with more than 80 researchers. Caveat: they see only what companies already disclose publicly.
States write AI rules and rely on companies to report their own results.
A state is paying for its own capacity to check those results.
Why it matters
Australia said yesterday it will compel outside evaluation of frontier models. New York is buying the cheaper half: evaluators of its own, with no power to compel anything. Whether published disclosures are enough to test a model is now a question two governments are answering differently.
Established
- Hochul announced a state programme to independently test AI systems.
- It runs through a Binghamton centre funded by $30 million and $25 million.
- Jeannette Wing and Greg Morrisett lead more than 80 researchers.
Not established
- Any power to compel access. Researchers see public disclosures only.
- A testing budget, or which models will be tested.
- That findings will be published or used in enforcement.
The five-minute readThe record, the timeline, the weight and 3 sources
The record
Reported, not verified. We read governor.ny.gov/news, where the newest releases are dated November 9, and the governor's technology pages. The only primary release on this centre is dated January 21, 2026 and announces its creation, not a testing programme. Politico had the launch first. We will verify if New York publishes a release naming the programme by December 10, 2026.
Politico reported on November 10, 2026, in an item timed at 13:00 Universal Time and carried by Yahoo, that Governor Kathy Hochul launched what the state calls a first in the nation effort for New York to independently test AI systems and to train AI safety specialists. The work runs through the Center for AI Responsibility and Research at Binghamton University, part of the State University of New York. Hochul is quoted: "The stakes are too high to simply take AI companies at their word." The centre is led by Jeannette Wing, executive vice president for research at Columbia University, and Greg Morrisett, vice provost at Cornell Tech, with more than 80 scientists and researchers from across New York. Funding is a $30 million gift from Tom Secunda, a Binghamton graduate and co-founder of Bloomberg LP, and a $25 million research investment from the State University of New York and New York State. Secunda chairs the board of Empire AI, the $500 million consortium that supplies the centre's computing. The report says researchers will see only what AI companies are required to disclose publicly, and that models still in development stay private. Tom Secunda is quoted saying the centre can test systems independently and give public officials evidence they can use. The launch is timed before New York's RAISE Act takes effect on January 1, 2027, which requires large model developers to publish safety frameworks that the governor's office hopes these researchers will examine. We read the January 21, 2026 release on governor.ny.gov that created the centre: it names the $30 million gift and the $25 million investment, announces a national search for leadership, and does not mention testing programmes or the RAISE Act. Not established anywhere we could find: a budget line for testing, which models will be tested, whether results will be published, and whether findings feed enforcement.
How it got here
- Jan 21 2026
Hochul announces the Center for AI Responsibility and Research at Binghamton, funded by a $30 million gift and a $25 million state and university investment, with a national search for its leadership.
- Sep 2026
California creates the first system of independent AI auditors in the United States. Edition 017 carded it as the first time a government put outside testers between a laboratory and the public.
- Nov 10
Australia publishes 6 mandatory obligations for frontier AI companies, one of them independent evaluation of their models, enforced by a proposed new regulator. Edition 084, this morning.
- Nov 10
New York launches the state testing programme, led by Jeannette Wing and Greg Morrisett with more than 80 researchers, 7 weeks before the RAISE Act takes effect.
The large model developers who must publish safety frameworks in New York from January 1; the federal Center for AI Standards and Innovation, which has published no safety report since August; Britain's and Australia's safety institutes; other state attorneys general; and the researchers who will discover how little public disclosure supports.
What to watch
- Would strengthen itThe New York centre publishes an evaluation of a named frontier model, or of a safety framework filed under the RAISE Act, by June 30, 2027.
- Would weaken itNo American state funds a comparable independent testing body, and New York's centre publishes no model evaluation, by December 31, 2027.
- Would strengthen itNew York or another state obtains access beyond public disclosure, by agreement or by law, to test a frontier model, by December 31, 2027.
Why it weighs this much
Cost is medium. $55 million of gift and state money is real and already committed, but most of it was announced in January and this is its deployment rather than new appropriation. Commitment is medium: naming 2 leaders and more than 80 researchers is harder to unwind than a press release, and the January gift binds the university, not the state. Reversibility is moderate. A governor can quietly stop citing a research centre, and nothing here obliges any company to do anything. Persistence is years: the RAISE Act takes effect on January 1, 2027, and an evaluator with standing computing access and a faculty tends to outlast the administration that funded it.
The words
- RAISE Act
- New York's frontier AI law, effective January 1, 2027. It requires large model developers to publish safety protocols and report serious incidents.
- Empire AI
- A New York public and private consortium, announced at $500 million, that supplies shared computing power to state universities.
- Independent evaluation
- Testing of a model by someone other than its maker. Its value depends entirely on what access the tester is given.
Sources
- GGovernor of New York, Jan 21, 2026: Hochul launches the nation's first independent AI research center at a public university, SUNY Binghamtongovernor.ny.gov
- YPolitico via Yahoo: New York's Hochul launches the nation's first state AI testing program, Nov 10, 2026yahoo.com
- WWiley: New York finalizes the RAISE Act for frontier AI models, law takes effect January 1, 2027wiley.law
Your call
Which explanation do you believe?
A call is final. Other people's calls show after yours.