LLM (google/gemini-3.5-flash-lite) summary:
- TECHNOLOGY BAN: new york city announced a sweeping ban on generative artificial intelligence for students through eighth grade
- ATTENTION SPANS: screens and social media have eroded the attention spans of students and teachers over time
- AI TUTORS: the ban improperly covers artificial intelligence tutors which can help adapt to individual student needs
- DIFFERENTIATION CHALLENGES: traditional classrooms struggle to address wide disparities in student skill levels without technological assistance
- COMPETENCY BASED LEARNING: artificial intelligence tutoring can help students progress at their own pace in subjects like math and science
- BALANCED APPROACH: a balanced school day could split time equally between nontechnology skills and maximalist technology use
- POLICY IMPLEMENTATION: the new york city policy announcement lacked a clear implementation plan and was released extremely late
- SYSTEMIC INEFFICIENCY: large urban school systems spend heavily per student while producing mediocre results and micromanaging educators
By
Ravi Gupta
Sept. 11, 2026 4:07 pm ET
I was a charter-school principal and superintendent in the South. I left that work a decade ago and have watched with alarm as screens and social media have eroded the attention spans of students and teachers, and large language models have spread faster than educators can adapt. So when New York City announced a sweeping ban on generative AI for students through eighth grade, I initially welcomed it.
The impulse to limit technology in schools is admirable. Childhood shouldn’t be an endless scroll. When AI may be able to do everything humans care about better than we can within 10 years, as Anthropic board member Reed Hastings recently said, it’s even more important to cultivate tactile skills and qualities of character that will endure when the labor market—and society more generally—is unrecognizable. Kids need to learn to use these technologies effectively, and I am sure they will, no matter what happens in schools. They don’t need us to teach them how to prompt a chatbot.
But when I dug into the details, I realized the AI ban goes too far. As of now, it covers the use of AI tutors, and that is shortsighted. Evidence for the effectiveness of AI tutors comes from a randomized trial of the Khanmigo program in Tennessee middle schools. That study found only modest improvement for the AI-tutored group. Yet a closer look at that study revealed that students, when left on their own, barely used the tutor.
The tutors—several on the market are reported to be effective—are amazingly effective at some of the thorniest pedagogical challenges, especially adapting to the individual needs of students. Educators call this “differentiation,” and it’s nearly impossible to do well without technology unless you spend enormous sums of money to get student-tutor ratios down to the single digits.
At my first school, Nashville Prep, we had kids entering fifth grade who knew algebra and others who could barely add. It’s impossible for one teacher, or even two, to give a class of 25 kids spread along that continuum what each one needs. Absent technology, we lean on “tracked” classes grouped by ability—a widely criticized practice because it can cluster disadvantage—or pick a midpoint and teach to the “average” student, a statistical fiction that ignores what actual students need.
So what’s the answer? I believe we should use AI tutoring to give students a path toward “competency-based learning.” That’s a fancy way of saying students progress at their own pace—something that is easiest in math and science. We should then train adults to motivate, coach and hold students accountable for their progress.
We need to be careful, however, not to keep students in front of screens all day. If I were starting a new school, I’d make it 50/50: half the day would involve no technology, the other half would be maximalist on technology, including AI tutoring and learning to use large language models for meaningful work. The nontech time would be used to read books, work through problems by hand, learn trades and other tactile skills, exercise, and play with friends.
Within New York’s announcement were a few tells that its leaders don’t understand the reality of schools. For one, they announced this change extremely late and without a clear implementation plan. One teacher wrote on X: “Are we still using I-Ready? Assessments still digital? What do STEAM classes look like now? Can we still do remote instruction? Where’s the list of banned/allowed resources so we can plan? School starts in 6 days.”
The city also announced a limited number of pilots of tech-based curriculum, evidence of the suffocating degree to which big-city school systems micromanage their educators. But teachers and principals don’t need a cue from on high to experiment, and they shouldn’t have to fill out reports every time they want to try something new. Strong educators are tinkerers, and we should let them experiment without asking for approval. By the time the approval comes, the technology and curriculum will inevitably have evolved enough to make the trial obsolete.
New York City spends more per student than any large school system in the country, and it produces mediocre results. It has been losing students—a trend that is expected to continue. So we should welcome AI tutors that could help reach the most vulnerable students and deliver high-quality instruction without unsustainable spending.
But AI tutors don’t pay union dues. That means powerful interests will be working against them. Those interests achieved a one-year reprieve, but I suspect they won’t be able to hold off these innovations for long.
Mr. Gupta is founder and former CEO of RePublic Schools and author of “The Invisible List: One Passion a Year, Every Year,” forthcoming Sept. 22.
Copyright ©2026 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8
Appeared in the September 12, 2026, print edition as 'New York City’s School AI Ban Goes Too Far'.






