New York City just released its new policy on AI and screen time in schools, and it's worth digging into what it actually says, because the headline version ("NYC finds the balance on AI in schools") isn't quite what's in the document.

Some quick history first. Back in January 2023, NYC schools banned ChatGPT outright, worried it would turn every essay assignment into a copy-paste exercise. That didn't last long. By spring, Chancellor Banks reversed course, and pretty bluntly too:

"The knee-jerk fear and risk overlooked the potential of generative AI to support students and teachers."

From there, without much citywide guidance, individual schools mostly set their own rules. Some leaned in, some didn't, and there wasn't a lot of consistency.

Now, under Mayor Mamdani and Chancellor Samuels, there's a citywide policy again, and generative AI is banned for students from Pre-K through 8th grade. Not "restricted." Banned. Chatbots, AI tutors, all of it, off for roughly 600,000 kids, with AI features being switched off across 38 existing software contracts to make it stick. Screen time gets capped too: no individual device use at all in Pre-K through 2nd grade, 30 minutes a day in grades 3-5, 45 minutes a day in middle school. High schoolers get something different: a handful of approved pilot programs in specific subjects, career-tech uses, and a cap of about 50,000 students citywide. Everyone else waits.

So let's be honest about what this is. It's not the "kids get access, with guardrails" policy I'd have written. For the vast majority of NYC public school students, this is a return to the first extreme, not a landing point between the two. I've said before that these tools, used well, can have a real positive impact on learning, that they can democratize education in ways we haven't had before, and that fluency with them is going to matter for these kids' careers whether we like it or not. None of that changes because a policy document says "moratorium."

There's a more immediate problem too. Kids are going to use AI whether school allows it or not, on their phones, at home, on tools nobody at DOE vetted. A classroom ban doesn't touch any of that. What it gives up is the chance to actually teach them how to use these tools well: what to trust, what to question, how to check a source, how to notice when a chatbot is confidently wrong. That's exactly the kind of critical thinking school exists to build, it was part of the justification for the "unban" in 2023, and it's harder to achieve once the tool is banned from the room where you'd be teaching it. Keeping AI out of the classroom doesn't make that need go away. If anything it makes things worse, because now the one place those conversations were supposed to happen isn't having them.

And yet, to be fair, I don't think this is the same mistake as 2023. This time it's explicitly framed as a one-year moratorium, not a permanent stance, with a Technology in Schools Coalition set to report back after this school year on what they've learned. There's a real, if narrow, high school pilot running in parallel, actually being studied, rather than a flat "no" everywhere. Teachers can still use AI for planning and operational work. That's a materially different posture than the 2023 ban, which was just fear with no plan attached.

One thing the policy does get right, and this must not get lost in the moratorium headlines: it preserves assistive technology access for students with disabilities. Screens and adaptive tech have been a genuine lifeline for kids who've always struggled in a traditional classroom, kids with special needs, nonverbal kids who found a way to communicate through a device long before anyone was talking about AI. These are exactly the students a blanket "turn it all off" policy would hurt most, and exactly the students who can least afford to wait for a task force to finish its report. Whatever changes as the pilots run and the review plays out, keeping their access intact has to stay non-negotiable.

All that said, if the choice were between a district that bans first and studies its way toward access versus one that opens the doors first and studies its way toward regret (especially for the youngest kids, where the screen-time argument stands on its own regardless of what's on the screen), then yes, the former is obviously preferable. But I reject the simplistic extremes, neither of those options strike the right balance which requires building the guardrails-and-access policy directly for every grade, studying the results, and learning and adapting in real-time.

Call this what it is: a cautious, partial course-correction, weighted heavily toward restriction for anyone under 14, with a real opening only at the high school level. It's not the balance NYC needs long-term. But if the review process is genuine, and the pilot data actually shapes next year's policy instead of getting shelved, this could be the first step toward one. That's a big if. I'll be watching what the Technology in Schools Coalition actually publishes, and so should you.