New York City's Schools Can't Finish an AI Policy. So It Stopped Buying.

New York City runs the largest school system in the United States, serving roughly 900,000 students. In March 2026, its Department of Education put out draft AI guidance built around a traffic-light framework: green for brainstorming and lesson planning, amber for some classroom uses, red for grading and discipline. By early July, the department had not managed to finalise it, and the schools chancellor told principals to stop buying AI software altogether until it does.

That is not a small district fumbling a rollout. It is the system every other superintendent watches, and it has spent five months failing to land guidance that its own community will accept.

What actually happened

According to Chalkbeat's reporting, the March draft drew nearly 6,500 public comments, most of them critical, and became the flashpoint for a wider revolt. In May, a marathon Panel for Educational Policy meeting saw parents and students demand a full moratorium. By June, First Deputy Chancellor Danielle Giunta was telling a City Council hearing that the department was pushing its promised June release back, citing the shifting national conversation around AI in schools.

More than half the City Council then signed a letter to Mayor Zohran Mamdani and Chancellor Kamar Samuels asking for AI use in schools to be paused outright, citing concerns about learning and student mental health. A separate petition for a two-year moratorium gathered thousands of signatures, and on 29 July parents and teachers rallied outside City Hall to press the case directly.

Samuels did not grant the moratorium. Instead, in a Monday email to principals reported by Chalkbeat and Gothamist, he asked schools to hold off on new software purchases until the department's updated guidance on AI and screen use is finally published later this summer. It is a procurement freeze, not a ban on AI already in use, a narrower, more administratively convenient move than what campaigners asked for.

Officials have now delayed final guidance twice while trying to write a policy specific enough to be useful and vague enough to survive public comment. Neither draft has managed both.

Why a New York procurement pause matters elsewhere

The timing is not incidental. A Christian Science Monitor investigation published 7 August described an absolute gold rush in district AI spending, noting the K-12 AI market grew to roughly $730 million in 2026 and is projected to reach $18.5 billion by 2036 on current growth rates. Part of that rush is manufactured: Anthropic and OpenAI have both offered free access to Claude and ChatGPT for educators, but only temporarily. Anthropic's offer is due to expire in June 2027, OpenAI's in June 2028. One expert quoted in the piece put it plainly: the vendors want to capture the market, a pattern seen with free tools in other sectors before pricing tightens once switching costs are high.

That is the backdrop against which NYC's pause should be read. A purchasing freeze in the country's largest district sends a market signal that a general note of caution does not. It also exposes a structural problem other districts are quietly wrestling with: many schools adopted tools during the free-access window without settled policy on grading, discipline or data use behind them, meaning the contract terms schools accept now will outlast the free period that made adoption easy.

The traffic-light framework, and why it broke

The substance of the March draft is worth studying regardless of what NYC ultimately publishes. A framework that greenlights brainstorming and lesson planning while red-lining grading and discipline is a reasonable starting position, and it tracks where the evidence on AI reliability is strongest and weakest. But it collapsed under two pressures that will recur anywhere a large system tries this: parents and unions read amber categories as a licence for drift into grading-adjacent uses, and staff without training had no consistent way to apply the categories day to day. A framework is only as good as the guidance and monitoring that sits underneath it, and NYC has repeatedly acknowledged it does not yet have that in place.

What to do

School leaders do not need to wait for a state or city framework to act locally. Put purchasing decisions on hold if a vendor contract outlasts any free-access period, and ask explicitly what the renewal price will be. Write down, in plain terms staff can actually apply, what AI may never touch, since grading and disciplinary decisions are the clearest lines emerging across the debate. Treat public comment periods as real signal: NYC's 6,500 responses were not noise, they were the policy failing in real time, and other districts drafting guidance should budget time to actually revise after consultation rather than publish and defend.

What to watch

Watch whether NYC's delayed guidance, whenever it lands, resolves the amber-category ambiguity that sank the March draft. Watch whether other large districts follow with their own purchasing pauses as free-access windows from major vendors start to look less like generosity and more like a runway to paid contracts. And watch the City Council's appetite for legislating a moratorium if the department's next draft fares no better than the last.

Sources: Chalkbeat: NYC delays school AI rules after criticism of draft guidance, Chalkbeat: Kamar Samuels asks NYC schools to pause software purchases, Chalkbeat: Parents demand AI moratorium at marathon PEP meeting, Christian Science Monitor: Education's AI gold rush comes with risks for students, K-12 Dive: NYC schools face public pressure to pause AI use for two years