NYC and LA Banned AI in Schools. That's Not the Fix.
A ban buys a year. It doesn't teach a child what to do with AI once the year is over.
On September 2nd, New York City announced a one-year moratorium on student-facing generative AI for kindergarten through eighth grade, roughly 600,000 students, nearly two-thirds of the district. The same week, Los Angeles Unified, the country's second-largest school system, went further: a blanket moratorium on generative AI across every student-issued device, every grade, for the entire 2026-27 school year. A day later, the American Psychological Association published a report that, read closely, explains exactly why a ban was never going to be enough.
What New York and LA actually decided
New York's policy isn't a total blackout. Elementary and middle schoolers lose open access to chatbots and AI writing tools in class, but high schoolers get twice-yearly “AI critical thinking” modules and a handful of supervised pilot classrooms. Los Angeles drew a simpler line: no generative AI on any student device, at any grade, for the whole school year. According to EdSurge's own reporting on the split, three of the five largest districts in the country, Chicago, Clark County, and Miami-Dade, took the opposite approach, approving specific, reviewed tools for classroom use instead of shutting the door on all of them.
The report that landed the next day
The APA's report wasn't about NYC or LA specifically. It was a broader warning that engagement with ed-tech isn't the same as learning, and it singled out generative AI as the clearest version of that problem: a tool that can improve a student's finished work without improving the student at all. Nicole Barnes, the APA's executive lead psychologist for education, put it directly, AI can help a student produce a better essay, but the improved essay doesn't mean the student became a better writer. That's not an argument for or against a ban. It's a precise description of the actual failure mode, one a moratorium doesn't touch.
The gap a ban doesn't close
A one-year pause buys time. It doesn't answer what happens when it lifts. A recent IBM-commissioned survey found generative AI is already used at least weekly in 76% of middle school classrooms and 73% of high school classrooms, while only one in five K-12 educators report having had any real training in how to use it responsibly. That's the actual shape of the problem: not too much exposure to AI, but too little structure around when a student is allowed to reach for it, and what they're expected to check before they trust what it hands back.
- A ban removes AI from the room; it doesn't teach a student what a right AI answer looks like versus a wrong one
- A twice-yearly module is a lecture, not a skill built through repetition
- A student who never touches AI under supervision meets it unsupervised the moment the ban lifts
- The APA's own finding is about sequencing: a tool can improve performance before it improves understanding, unless the two are deliberately kept in step
“Banning AI for a year doesn't make a student ready for the year after. The question was never whether a child uses it, it's whether they've built enough of the fundamentals by hand to know when it's wrong.”
Muhammad Mostafizur Rahman, Founder, PsyntaxLabs Academy
What actually builds that judgement
PsyntaxLabs Academy doesn't hand a student AI tools on day one either. Across the 24-week programme's four stages, Use, Build, Product, Launch, AI only enters the picture once a student has already built the fundamentals of the thing by hand: working code, structured logic, decisions they can defend line by line. By the time AI is in the loop, an answer it hands back is something to check, not something to accept, because the student already knows what a right one looks like. That's the same sequencing the APA's report is describing, just applied deliberately instead of left to chance, and cohorts capped at six mean a founder-mentor is actually there to catch the moment a student stops verifying and starts trusting blindly.
New York and LA are right that unstructured AI access in a classroom is a real problem. They're just treating the exposure as the problem instead of the missing structure around it, the exact distinction the APA's own report draws. If you want your child using AI the way that research says actually works, fundamentals first, so they can tell a right answer from a wrong one, book a free consultation. We'll show you exactly where AI enters our curriculum, and why not sooner.