Hochul says Stanford AI reviewed every New York rule in months — flagging a $25 dog-hunting fee — in work she says would have taken staff five years

New York just used AI to do something state governments are not built to do quickly: read their own rulebook. Governor Kathy Hochul told Bloomberg's Odd Lots podcast that her team fed the state's full body of laws into a Stanford RegLab system and, with human reviewers in the loop, surfaced thousands of streamlining opportunities in a couple of months. The work, she said, "probably would have taken five years at the staff level."

The speedup is the story, not the list of weird old laws the AI flagged, though the list is what catches a reader. The governor's office has cited examples: a $25 fee to take a dog hunting, and a stipulation that pregnant people need a permit to work past midnight. Both are the kind of dead-letter rules that accumulate in any state that has been writing statutes since the 19th century and rarely deletes them. The point of the review is to surface them. What happens next is a normal legislative and regulatory process.

That distinction matters because the speed claim is easy to misread. AI did not pass a new law; it did not strike the dog-hunting fee. It did the part humans are slow at: reading everything. The five-years-to-two-months comparison is real for the surfacing step. Whether any of the surfaced items becomes an actual repeal still depends on the same machinery that the five-year timeline was about to begin. As Hochul put it on the podcast: "I want a government that's not on your back but on your side, and using AI has been powerful to do that... I'm going to make dramatic changes using the power of AI." The framing is constructive, a public tool being used to clean up public rules, not "AI solved government."

The same news cycle produced a separate, differently motivated move. New York became the first state to pause new hyperscale data centers for up to a year while it drafts rules around utility costs and environmental impact. "Hyperscale" here means the largest AI-grade computing facilities, the kind that draw hundreds of megawatts and put real load on a regional grid. The pause is about siting and infrastructure, not about whether AI should be used inside state government. Treating it as a contradiction with the rule-review exercise requires reading the governor as broadly "for" or "against" AI, which the source basis does not support. The two moves are running on different tracks: one is administrative self-audit accelerated by machine reading, the other is industrial-policy friction over power and water.

The mechanism behind the rule-review story is a structural mismatch, not a one-off efficiency win. State rulebooks grow faster than staff can audit them, and the cost of that gap is measured in the dead-letter fees and midnight-work permits the public only notices when one of them hits. If a Stanford-built system can compress the read step from years to months, and if a governor is willing to publish the result, the bottleneck shifts from finding the antiquated rules to deciding what to do with them. The governor's office says the recommendations are "supplemented by human review." The speedup is the surface, not the enactment.

The watch items sit there. The review produces a list, not a statute. If the surfaced streamlining opportunities stall in committee or get quietly dropped, the "months not years" claim ages badly, because the speedup was always in surfacing, not in repealing. The data-center pause has its own clock: the state is supposed to draft the new siting rules within the year, and the result will tell the industry whether New York is a long-term host or a long-term holdout. Hochul's "dramatic changes" framing is forward-looking. The honest way to read it is as a commitment, not a delivery.