How city governments, including New York, are using artificial intelligence, and where things are going next
It seems like both the people and the politicians in every American city are talking about AI. Far fewer have figured out concretely what to do with increasingly powerful automated technological tools.
New York is one of a few cities that publishes exactly which algorithms its agencies use. Under Local Law 35, which was passed in 2022, just before ChatGPT burst into the marketplace, every agency must report each year on tools that use machine learning or predictive analytics to make or support decisions that affect the public.
The City’s latest report covers 2025, listing 86 tools at 19 agencies — an increase from 31 tools at 11 agencies in 2022.
What do they do? A range of things. The Administration for Children’s Services uses an algorithm to rank open child-abuse investigations by the predicted risk of severe harm and decide which ones get extra review. The Fire Department scores about 330,000 buildings for fire risk, factoring in past fires, building characteristics, 311 complaints and nearby felony crimes. Then there are more basic applications many of us take for granted in other contexts: caption generators in Zoom and Adobe, ChatGPT for drafting social media posts and so on.
In some cases, the City has also put AI directly in front of New Yorkers, with results that can fairly be called mixed. The school system’s help-desk chatbot, called Eureka, fields about 1,500 calls and more than 300 online inquiries a day — and, according to the City, resolves about 30% of the calls and 10% of queries. The Department of Youth and Community Development’s chatbot fields about 2,100 questions a day, delivering preset answers to a range of questions. And the 311 app now lets users snap a photo of a problem and get a suggested complaint type.
Not all of these are sophisticated uses of AI, but the umbrella term is broad enough to encompass many types of software.
The highest-profile experiment went belly up. Under Mayor Eric Adams, the City launched its MyCity business chatbot, and before long, reporters at The Markup and elsewhere found that it was advising people to break the law — for instance, telling business owners they could take a cut of workers’ tips and telling landlords they could turn away tenants with housing vouchers. Mayor Zohran Mamdani shut it down early this year.
Mamdani has said he doesn’t personally use ChatGPT or Claude. The same cannot be said for the government he oversees, and experimentation with AI is likely to grow as it gets less expensive and more capable.
The hard calls
Policing
There are many potential AI uses in public safety — too many to list. Some are relatively uncontroversial; others clearly raise surveillance and privacy concerns.
Many fall in the in-between. Facial recognition shows up three times on the City’s list: at the NYPD, which has used it since 2011, and at the Department of Investigation and the Fire Department's fire marshals.
The NYPD describes its use of the technology as narrow. Investigators compare a still from surveillance video with the department’s pool of arrest and parole photos. Searches beyond that pool are supposed to get sign-off from top officials, and the NYPD — in a statement after a wrongful arrest facilitated by a bad match — said officers “cannot and will never make an arrest solely using” the technology. Clearview AI, a controversial face-recognition company that built its database by scraping photos from social media and the open web, is officially off-limits to cops (after some unofficial use was revealed).
The NYPD also reports using a machine-learning tool called Patternizr, which flags burglaries, robberies and larcenies that may be linked — and ShotSpotter, the gunshot-detection system, which the City counts as an algorithmic tool because software classifies sounds as potential gunfire before human analysts review them.
Separately, a law called the Public Oversight of Surveillance Technology (POST) Act, passed in 2020, requires the department to publish an impact-and-use policy for each surveillance technology it uses. The reporting page for the law lists dozens, from license plate readers to drones and robots.
Education
In recent years, the city’s schools, teachers and students have experienced pendulum swings on AI. The Education Department banned ChatGPT in 2023, then lifted the ban months later. This year, Chancellor Kamar Samuels proposed an AI-themed high school — then killed the idea after criticism.
May 2026 system-wide AI guidance allowed some uses of AI, including for lesson-planning by teachers (but not for grading); then, in September 2026, Mamdani and Chancellor Kamar Samuels announced a more restrictive policy, touting “the nation’s broadest generative AI moratorium in schools.”
For now, that bans any student use of chatbots like ChatGPT and Gemini that produce text, images and other content from 3-K through eighth grade. High schoolers are set to get limited exposure through a handful of supervised pilots.
The zig-zagging is unsurprising. Districts all over the country are struggling to figure out how teachers and students should approach the technology. In a 2025 RAND survey, 54% of students said they used AI for school, but only 45% of principals said their school or district had any policy on it.
Cost-cutting and the workforce
Should AI be used to shrink the size of government and increase efficiency? Perhaps. So far, Mamdani’s City Hall has said AI will not come “at the expense of workers, our environment or our civil rights.” In theory, that leaves the door open to using AI to streamline and speed burdensome processes.
In July, Mamdani announced five Public Interest Technology crews inside the Office of Technology and Innovation. While the crews aren’t using AI, at least at first, one built a portal for complaints about subscriptions that are hard to cancel in about 10 weeks. An OTI official said that saved months of procurement time.
If the City pushes on this front, negotiations with public-sector unions may set the limits of how far AI can be used to improve operations and potentially replace human tasks and workers.
State law sets limits, too. The LOADinG Act, which Gov. Kathy Hochul signed in 2024 and which has since been extended to cities, bars public employers from using AI in ways that displace workers, cut their hours, wages or benefits or hand their existing duties to automated systems.
How AI is reshaping the city itself
Government is being transformed at a far slower pace than the private sector. The Census Bureau’s regular business survey now asks firms whether they used AI in the past two weeks. About 17% of New York businesses said they had in early April 2026, a bit below the 20% national rate.
Other data suggest broader use. The New York Fed, which asks whether firms used AI at any point in the previous six months, found that 61% of service firms in the New York-Northern New Jersey region did in 2026, up from 25% in 2024.
Fears of AI-fueled job losses are common, and those losses could be coming. But so far, just 4% of the region’s service firms told the New York Fed they had laid anyone off because of AI, while about a third of those using it said they had retrained workers.
Many recent college graduates in the city (and nationally) are having a difficult time in the job market, and some suspect AI as a culprit. In the year ending in March 2026, New Yorkers 22 to 27 with a degree had a slightly higher unemployment rate (7.3%) than those without one (7.1%), the first time on record that has happened, according to the City Comptroller.
Meanwhile, New York, which has trailed San Francisco in the tech sector for years, sits near the top of the country’s AI economy: Brookings ranks the metro among 28 “Star Hubs,” the tier just below the Bay Area, on AI talent, research and business adoption.
While other parts of the economy feel the impact of the advancing technologies, AI companies are also filling Manhattan offices. OpenAI, Anthropic and their rivals have leased about 4 million square feet in the city since 2019, accounting for 32% of tech-company office demand at the end of 2025, according to the brokerage company CBRE.
What do these forces mean for New York City’s famously resilient economy and the city’s budget? In May, City Comptroller Mark Levine published the first attempt to model scenarios, assigning probabilities to five possible outcomes. The one deemed likeliest, at 35%, is an “AI-Empowered Economy” in which productivity rises with little disruption. The others are an AI investment bust (25%), widespread job replacement (20%), a productivity boom (15%) and an “AI Shockwave” (5%) in which white-collar jobs vanish faster than the labor market can absorb displaced workers.
The probabilities might sound like guesswork, and that’s because no one really knows what’s next.
What to do about it all? Right now, much of the focus is on regulating AI risks amid fears that AI-empowered hackers and AI agents run amok will do damage that can’t be contained.
At a City Council hearing on Oct. 5 local lawmakers grilled company leaders and discussed possible new local regulations — even as many critics cautioned that setting guardrails should be a state or federal responsibility. Already, there are some regulations on the books here: Local Law 144 requires employers that use automated hiring tools to commission annual bias audits and tell applicants. (It doesn’t seem to be working very well: A December 2025 audit by State Comptroller Thomas DiNapoli found the enforcement by the Department of Consumer and Worker Protection to be “ineffective.”)
In other parts of the country, data centers and Flock license plate cameras have become major campaign issues. Here in the city, concern is growing on other fronts. Whether any of this will shape future adoption of AI by individuals, businesses and government is anyone’s guess.






