New York launches state-run testing of AI models
The program aims to replace company self-reporting with independent checks, but it can only examine what developers disclose publicly.
New York has launched what it calls the first state-run program for testing AI models, aiming to have independent researchers verify AI companies' claims instead of relying on the companies to police themselves. The work will be carried out by the New York Center for AI Responsibility and Research at Binghamton University, with more than 80 state scientists drawing on computing resources from Empire AI, a public-private partnership backed by $500 million. The center is tasked with assessing advanced systems, examining their risks and supplying evidence to guide officials, including on responsible uses in medicine and scientific research. A limitation is that the program can only examine what developers disclose publicly. Under the state's RAISE Act, large developers must register with New York by January 1, 2027, after which they face obligations to release safety frameworks, transparency reports and incident reports on a short deadline, though the law does not reach models that remain unreleased.
Why it matters
For readers following AI policy, the shift is from taking developers at their word to having state scientists test their claims. The limit is that this scrutiny depends on what companies choose to publish, so unreleased models fall outside it. Under the RAISE Act, large developers must register by January 1, 2027, after which they face disclosure duties, but those duties also stop short of unreleased systems. Together, the setup gives the state a check on public claims while leaving a gap before models reach the public.
Matters or hype?
Does this story matter, or is it hype? Vote before you see what everyone else said.