Key Takeaways
- New York City’s ban on generative AI for students in grades K‑8 (with limited high‑school exceptions) was issued without addressing how different AI tools function or what data they collect.
- The policy treats all “AI” as a single category, sweeping together chatbots, scoring engines, and other technologies that serve distinct purposes in education.
- Teachers already spend unpaid hours after school analyzing student work; banning useful teacher‑facing tools removes the support that makes rigorous, shared curricula effective.
- Evidence from a district pilot showed that a teacher‑facing scoring tool raised student performance by up to 16 % and lifted district scores by roughly 11 points on state tests.
- The author argues that regulations should focus on the function of AI tools, require proof of efficacy, and set clear timelines for decision‑making, rather than relying on blanket bans.
- Effective governance in education demands that leaders first understand the technology and then communicate its implications to the public—something the city has yet to do for AI in schools.
The Reality of After‑School Teaching Work
At 9 p.m. on a Tuesday night in the Bronx, a seventh grade math teacher sits at her kitchen table with a stack of exit tickets. She isn’t assigning grades. She’s figuring out which of her 24 students missed the same thing today, and why, so tomorrow starts better. That expert work is done alone, unpaid, at the end of a 14‑hour day. The author recalls doing the same for six years in the South Bronx, noting that despite early warnings, the bulk of teaching happens after the bell rings.
NYC’s Generative‑AI Ban and Its Scope
Schools Chancellor Kamar Samuels and Mayor Mamdani have announced that generative AI is banned for students from 2‑K through eighth grade, with narrow room in high school for AI literacy and career programs. Teachers may still use approved tools to plan and handle operations, but not for grading, IEPs, or behavior tracking. Multilingual learners and students with disabilities receive exceptions. The rule appears sweeping, yet it leaves many practical questions unanswered.
The Flawed Color‑Coded Guidance
Back in March, Samuels presented his draft guidance with a color‑coded chart—green, yellow, red—that never answered the real question: what happens to a child’s data and thinking. The chancellor admitted in May the city missed the mark, then banned the category instead of fixing it. As the author puts it, “Easier, but far less useful, because ‘AI’ isn’t one thing. A chatbot and a scoring engine are as different as a slot machine and a stethoscope, and a rule at the word ‘AI’ can’t tell them apart. It just sweeps.” The policy conflates disparate technologies under a single label.
Curriculum Reform Meets After‑Bell Support
For four years, NYC Reads and NYC Solves have asked teachers to set aside materials built over entire careers for a shared, rigorous curriculum. That call is right, and it is the hardest ask there is, since curriculum doesn’t implement itself. The gap closes after the bell. Ban the tools built for that work and the mandate stays while the support disappears. Nobody set out to undercut this; the effect is what students live with.
Evidence from a District Pilot
Last year in District 11, nearly 6,000 middle schoolers and 180 teachers used a tool my company built to score daily work and point to the next lesson. Students whose teachers used it outperformed peers by up to 16 %, and District 11 gained roughly 11 points on the state test against a comparable district. Some of that is surely stronger teachers using better tools. It deserves scrutiny, not less of this. The results demonstrate that a well‑designed, teacher‑facing AI aid can lift achievement when integrated thoughtfully.
The Missed Opportunity to Engage Vendors
None of it required a fight. In July, my team showed the city, point by point, that Kiddom already met every requirement in its guidance. It also answered the exact privacy questions parents were raising. Neither got a reply. The specificity everyone says is missing already exists. The city’s silence suggests a reluctance to engage with concrete evidence rather than a genuine lack of information.
A Function‑Based Regulatory Approach
Here’s what I’d build instead: write the rule at the level of function, not the label, since student‑facing and teacher‑facing are different animals. Make evidence the price of entry, and any company in front of New York’s kids should publish its results. Most will fail. That’s the point. Put a date on it and reach a decision. By focusing on what a tool actually does—whether it analyses student work, provides feedback, or merely generates text—regulators can differentiate harmless chatbots from powerful instructional aids.
Leadership, Learning, and Public Communication
A democracy holds together when leaders learn the subject, then teach it to the people they serve. Skip the first and you get bad policy. Skip the second and even a good call looks like it fell from the sky. Mamdani has shown he can do that: a three‑minute video explained a pied‑à‑terre tax to millions. Nobody’s made that video about a scoring engine, or why it’s nothing like a chatbot. That’s the real failure, more than the ban. New York knows how to teach. It just hasn’t.
About the Author
Manjee, a former NYC public school math teacher, is the co‑founder and chief academic officer at Kiddom. His perspective blends classroom experience with the realities of ed‑tech development, informing his call for smarter, evidence‑based AI policy in New York City schools.