Mayor Zohran Mamdani and Schools Chancellor Kamar Samuels announced the nation's broadest school AI moratorium on Sept. 2. It pulls generative artificial intelligence from classrooms serving 600,000 New York City students, two-thirds of the system's enrollment. The ban runs through the 2026-27 school year1. High schools keep a narrow exception: pilot programs capped at 50,000 students and five classes per building1. The policy bars companion chatbots and every student-facing generative tool from pre-kindergarten through eighth grade1.
Chart 1
New York City bars generative AI from 600,000 students while admitting it to a 50,000-student pilot
New York City public-school students affected by the 2026-27 generative AI moratorium, by group, in thousands
Each column counts students in one group: the magenta column is students barred from generative AI in grades pre-kindergarten through eighth, and the gray column is the cap on the high-school pilot program.
Source: NYC Mayor's Office [1]. Chart by The AILately.com Desk.
The numbers behind this chart
| Item | Value | Note |
|---|---|---|
| Barred, pre-K–8 | 600,000 | Two-thirds of the system's enrollment |
| HS pilot cap | 50,000 | Five classes per building at most |
Read quickly, the order looks like courage, the nation's largest school system telling Silicon Valley to wait. Google turned on its Gemini chatbot inside Google Classroom for students of every age worldwide in August. The rollout reached more than 150 million users before most schools decided anything at all7. One ban in one city left the default switched on for the rest of the country's children that same season. The ban measures what New York City chose to refuse, leaving untouched what every other classroom in America just received, unasked.
Formal policy reaches only 18 percent of the country's K-12 teachers, a Gallup poll commissioned with the Walton Family Foundation found in May4. Forty-eight percent pick up informal guidance from whatever their building passes along, and the remaining 34 percent work from their own judgment alone4. Robin Lake, director of the Center on Reinventing Public Education, summed up the result to NPR: "It feels very Wild West in terms of the randomness"3. Teachers using AI tools still save close to six hours a week, the same poll found4. Most spend those hours as the only judge of what the tool told a student.
Training follows money before it follows evidence. Low-poverty districts offered teachers formal AI training at a 67 percent rate by fall 2024, up from 43 percent a year earlier. High-poverty districts reached 39 percent by the same measure, up from 6 percent, RAND Corporation survey data reported by K-12 Dive found5. The gap narrows with each year of data. It still separates a teacher who sat through a real training session from one who received a login and a shrug.
Chart 2
High-poverty districts triple AI training to 39 percent in a year, still 28 points behind low-poverty districts
Share of U.S. school districts offering teachers formal AI training, by district poverty level, fall 2023 and fall 2024
Each row is one poverty tier of districts: the blue dot marks fall 2023, the magenta dot marks fall 2024, and the rod between them is how far the share moved in a year.
Source: K-12 Dive [5]. Chart by The AILately.com Desk.
The numbers behind this chart
| Item | Fall 2023 | Fall 2024 |
|---|---|---|
| Low-poverty districts | 43% | 67% |
| High-poverty districts | 6% | 39% |
Dayton Public Schools paid $72,600 last year for an AI platform called SchoolAI, then renewed the contract this year for $78,6503. Salamanca City Central School District, in western New York, paid nearly $58,000 for a humanoid robot named Sally, built by Realbotix3. Stanford researchers reviewed more than 800 academic papers on AI in K-12 education this year. The review still could tell districts little about which of these purchases teach a child anything3. Dayton Superintendent David Lawrence defended the spending anyway: "I think when the history is written, it'll say that 'Hey, they were really cutting-edge, innovative and ahead of the curve in the interest of children,'" he told NPR3.
Lake makes the case for keeping the door open anyway. "If we really close off AI experimentation, if we prevent kids from learning about AI, learning how to use it really well, thinking about how teaching could be transformed, we will set our kids back from being able to navigate a world that is changing outside of schools faster than we can possibly imagine," she told NPR3. Her sentence runs long because the stakes she names run wide. She names experimentation, transformation, and a world that moves before policy catches up to it.
Samuels frames the retreat differently. "We're putting guardrails in place to protect the human connection, curiosity and creativity that help children grow," he said at the announcement1. Guardrails, in his words, protect something a chatbot lacks. Mamdani put the same point more plainly: "Children need teachers and human connection in order to learn and grow." He added, "Innovation does not mean more technology"1.
The equity question sits inside the ban itself. Jean-Claude Brizard, head of Digital Promise, made the opposing case within days of the announcement: "Kids of middle income, upper income parents are going to have access to technology. They're going to learn how to use AI," he said2. Families with money buy a tutor, a subscription, a laptop loaded with the tool at home. Two-thirds of New York City's public-school students, now barred from the same chatbot at school, depend on whatever version of that access their household can afford.
Schools faced a version of this choice before. In 2015, more than a third of lower-income households with school-age children lacked a broadband connection at home6. Pew Research Center measured that gap, and the Federal Communications Commission came to call it the homework gap. Districts spent the following decade closing it one grant at a time, after children who lacked the connection had already fallen behind in subjects that required it. AI training now follows the same shape: 67 percent against 39 percent, a gap drawn by a district's tax base before any lesson plan gets written.
Ask who pays. New York City's youngest students pay in the practice they miss, whichever side of the argument turns out right. Teachers nationwide carry the liability Gallup counted, 82 percent of them working the job alone. Google, SchoolAI and Realbotix collect the revenue and the user count, whichever city bans the product or whichever superintendent calls it cutting-edge. The risk lands hardest on the student whose district had the least capacity to write a policy either way, banned or open.
A ban photographs well. Training budgets draw fewer cameras and ask for real money every year. That gap in cost is most of the reason New York City could write a ban and Dayton settled for whatever a vendor was selling instead. New York City, with 600,000 students to protect, chose the cheaper kind of protection, a closed door for a year. A trained teacher in every room, regardless of a district's tax base, remains the costlier kind, the one still waiting on a budget line.
Readers who want the money and the training tracked district by district can join The Signal, AI Lately's daily email. It arrives each morning.
Sources
- Staff, "Mayor Mamdani and Chancellor Samuels Put Students First with Nation's Broadest Generative AI Moratorium in Schools," NYC Mayor's Office, Sept. 2, 2026, https://www.nyc.gov/mayors-office/news/2026/09/mayor-mamdani-and-chancellor-samuels-put-students-first-with-nat
- Alyson Klein, "New York, Los Angeles Restrict Student AI Use. Will Other Districts Follow?," Education Week, Sept. 2, 2026, https://www.edweek.org/technology/new-york-restricts-ai-in-elementary-and-middle-schools-will-other-districts-follow/2026/09
- Lee V. Gaines, "Schools Are Experimenting With AI With Little Evidence or Policy to Guide Them," NPR, Sept. 29, 2026, https://www.wfyi.org/education/2026-09-29/schools-are-experimenting-with-ai-with-little-evidence-or-policy-to-guide-them
- Andrea Malek Ash and Katherine Senseman, "Most Teachers Receive No Formal Guidance on AI Use," Gallup, May 26, 2026, https://news.gallup.com/poll/710534/teachers-receive-no-formal-guidance.aspx
- Anna Merod, "Teacher AI Training Remains Uneven Despite Uptick," K-12 Dive, April 18, 2025, https://www.k12dive.com/news/teacher-ai-training-remains-uneven-despite-uptick/745668/
- Emily A. Vogels, "Lower-Income Americans Still Less Likely to Have Home Broadband, Smartphone," Pew Research Center, June 22, 2021, https://www.pewresearch.org/short-reads/2021/06/22/digital-divide-persists-even-as-americans-with-lower-incomes-make-gains-in-tech-adoption/
- Grace Jiang, "Schools Caught Flat-Footed After Google Makes Gemini Chatbot Available to All Students," Education Week, Sept. 1, 2026, https://www.edweek.org/technology/schools-caught-flat-footed-after-google-makes-gemini-chatbot-available-to-all-students/2026/09