New York pulls AI from nearly 600,000 students

September 10, the backpacks come up light

Thursday, September 10, the first day back in New York. Close to 600,000 students, from 2K through 8th grade, will walk into class without a single piece of generative AI software. That's two-thirds of the enrollment in the country's largest school district.

Their teachers get to keep theirs.

That's the detail most of the coverage skipped, and it's the detail that actually explains the decision. Mayor Zohran Mamdani and Chancellor Kamar Samuels didn't pull AI out of schools. They pulled it out of kids' hands.

Mamdani framed it as a refusal, not a precaution: "The tech industry wants us to believe that A.I.-powered early education is not only inevitable, but necessary. We do not see it that way."

What the policy bans, and for whom

The guidance published by New York City Public Schools is short and doesn't hedge. Software built on student-facing generative AI isn't permitted in grades 2K through 8. It's a one-year moratorium, running through the 2026-2027 school year.

AI companion chatbots are banned at every grade level, high school included. The policy also caps 1:1 screen time: none at all for the youngest kids, then a recommended ceiling of 30 minutes a day in grades 3-5 and 45 minutes in grades 6-8.

High school keeps a narrow opening. Five tools have been approved district-wide, capped at 50,000 students total, roughly 5% of enrollment, no more than five classes per school, and one pilot program per student. Every high schooler will also take two mandatory 45-minute modules on what AI is, and where it's biased and risky. A coalition of students, parents, unions, and researchers will track the year and publish a report.

The five tools are named: Quill for English and reading, Edia for math, Brisk Teaching on texts the teacher selects, plus Playlab and Intel AI-Ready Schools across several subjects. Weekly doses published by the district range from 15 minutes to a full class period depending on the tool, always under the supervision of a trained teacher. Edia stays in the classroom, never homework.

There are carve-outs, and they're spelled out. Assistive technology required by a student's individualized education plan still applies, along with tools for English language learners and career and technical education programs. The guidance puts it plainly: "This policy limits unnecessary AI and screen use, not necessary use."

The line runs between preparing and judging

Here's the passage that matters, and it's two sentences in the guidance. Teachers and other school staff members may continue using AI "for instructional planning and operational tasks." Then, right after: "AI may not be used for grading, behavior monitoring, or for placement, promotion, graduation, and other decisions about students."

New York drew a moral line before it drew a technical one. The machine can help an adult build a lesson. It can't sit in judgment of a child. A teacher can ask it for three versions of a fractions worksheet; they can't ask it what Maya's essay is worth.

The guidance adds a line that reads like a warning shot: "Teachers remain responsible for the quality, accuracy, and appropriateness of what they produce. AI may assist educators, but it does not replace professional judgment." That's a rule about accountability, not about the tool. You can outsource the drafting. You can never outsource the signature.

The same district said the opposite three years ago

New York has run this scene before, in reverse. In early 2023, weeks after ChatGPT launched, the district added the site to its list of filtered pages. A few months later, then-Chancellor David Banks published an op-ed walking it back.

He called it "knee-jerk fear," said the restriction never really amounted to a ban, and pointed to a Queens middle school as his example. Its teachers, he wrote, were "experimenting with AI to create personalized lesson plans and grade papers."

Grade papers. That's the exact same move the 2026 guidance just outlawed, in the same district, three years later.

One nuance is worth keeping straight, because it changes what the story is. What happened in 2023 wasn't a ban: ChatGPT joined a list of filtered sites that also included YouTube, Netflix, and Roblox, and any school could request access. This is the first time New York has actually banned it.

What the policy leaves out

Three caveats matter, starting with how much legal weight the text actually carries. The screen-time caps are written as recommendations, not requirements. The guidance says it's "recommended" to stay under 30 and 45 minutes. Only the youngest students, up through 2nd grade, face a real ban on 1:1 screen work. A recommendation in a district this size is a speed-limit sign with no camera behind it.

Second, no tool has been named for removal. The guidance bans student-facing AI software and announces a broader review of every digital tool in the district, with the power to cut some loose. But that review is announced, not done. We know what's banned now. We don't yet know what gets unplugged.

Third, the five high school tools are described as "carefully vetted." The guidance itself says what that vetting actually covers: data privacy and security. It adds that the review doesn't yet extend to "algorithmic bias, equity impact, and instructional effectiveness." In other words: New York checked what these tools do with kids' data, not whether the tools make kids learn better.

Two systems, two answers

Three months ago, we covered Estonia's AI Leap program: 154 high schools wired up to an educational version of ChatGPT in five months, built on a stated doctrine, teach kids to argue with the machine instead of doing without it. Tallinn is equipping and training. New York is pulling back and giving itself a year to watch.

These are two opposite answers to the same question, and it's the only question that matters here: what happens to a school once effort becomes optional? We argued here that the friction taken away from a child was the learning itself, not busywork bolted on the side. A school system just wrote a policy on that premise.

State Senator Robert Jackson, who backed the guidance, put it his own way on the day it was announced: "We should never confuse a machine's ability to produce an answer with a child's ability to arrive at one. There is something sacred in the struggle to understand: the question asked twice, the page wrestled with, the mistake that becomes discovery." New York has given itself a year on the calendar to find out if he's right.