NYC Banned AI For Under-14s. England Is Betting the Opposite Way

On 2 September, New York City did something no large school district had done before: it banned generative AI outright for students from pre-kindergarten through eighth grade. Mayor Zohran Mamdani and Schools Chancellor Kamar Samuels announced a one-year moratorium covering roughly 600,000 pupils, nearly two-thirds of the city's public school enrolment, according to reporting from Chalkbeat, the Washington Post and CNN.

The policy prohibits chatbots, AI tutors and other student-facing generative tools for children up to around age 14. Companion chatbots that offer emotional or mental-health support are banned outright for all students, regardless of age. The city says it will "discontinue or disable the AI components of more than 38 previously allowed programs" that do not meet new safety and oversight standards. High schoolers keep access, but only under what officials call "restricted" rules, with pilot programmes still to be worked out.

There's a pointed history here. NYC banned ChatGPT in schools in early 2023, then reversed that decision within months, saying the initial fear had "overlooked the potential" of the technology. This new moratorium is a much larger and more deliberate reversal of that reversal, built on two years of pilot data, safety reviews and, notably, sustained pressure from the state teachers' union.

Why now, and why this shape

The New York State United Teachers board passed a resolution in May backing bans on student-facing AI in early grades and prohibiting chatbots for under-16s. Michael Mulgrew, head of the city's teachers' union, welcomed the accompanying screen-time limits but said the wider AI policy "still leaves many questions unanswered" — a sign this is a negotiated compromise, not a technologist's clean design.

A coalition pushing for a full two-year moratorium called the policy a "step in the right direction" but flagged concern that tens of thousands of high schoolers will still get generative AI access through pilot programmes. In other words, nobody who campaigned on this issue thinks the line is drawn in quite the right place — which is often what a genuine compromise looks like.

The policy does carve out exceptions: coding software, and tools that support students with disabilities or English language learners are exempt from the ban. That distinction matters. It suggests the moratorium is targeted at open-ended conversational AI and instructional chatbots, not at assistive technology more broadly.

The mirror-image bet in England

Compare this with where England has been heading. The Department for Education has added AI literacy to the curriculum, expanded testbed funding, and is piloting AI tutoring for hundreds of thousands of pupils on free school meals — moves aimed at pupils in the same age bracket NYC just locked out. Keeping Children Safe in Education 2026, which came into force on 1 September, treats AI-enabled harms as a mainstream safeguarding issue but does not restrict age-based classroom use the way NYC's moratorium does; it focuses on filtering, monitoring and deepfake-related risks rather than banning tools outright.

So two English-speaking systems, dealing with the same evidence base and largely the same vendors, have reached opposite starting positions for children under 14. NYC has decided the risk-to-benefit ratio doesn't clear the bar yet and is choosing to wait. England has decided the risk of pupils falling behind on AI fluency outweighs the risk of premature exposure, and is choosing to teach through it.

Neither system has strong causal evidence for its choice. Both are, in effect, running a one-year natural experiment on primary and lower-secondary pupils, at national scale, without a control group.

That's not a criticism unique to either city or country — it's the current state of the field. The OECD's own 2026 analysis found AI use lifts task scores in the short term while learning measures drop over time, which cuts against both a total ban (which forfeits any benefit) and unrestricted rollout (which risks the learning cost). NYC and England have each picked a side of that trade-off for younger pupils, for different political and industrial reasons.

What to do

School leaders in England should not assume the DfE's curriculum push settles the youngest-pupil question. If NYC's data over the next year shows measurable gains in engagement or safeguarding outcomes from removing generative AI for under-14s, expect that evidence to surface in UK policy debates quickly. Primary and lower-secondary leaders should keep a clear separation in their own AI policy between assistive tools (support for SEND, EAL) and open-ended generative chat, since that is exactly the line NYC drew.

What to watch

Watch for NYC's own review at the end of this school year — the moratorium is explicitly one year, not permanent — and for whether England's curriculum rollout produces any comparable evaluation data for the same age group. A genuine head-to-head comparison, even an imperfect one, is rare in this field and worth tracking closely.

Sources: Chalkbeat, Washington Post, CNN, NYC Mayor's Office, Computing at School on KCSIE 2026