What NYC Public Schools published

New York City's Education Department unveiled its preliminary guidelines for artificial intelligence use in March 2026, offering a road map for if and when to incorporate AI tools in school. The guidance arrived nearly three years after a short-lived school ban on ChatGPT, and in the midst of ongoing debates about student privacy, AI's effect on student learning and development, and the role of private companies in schools. A full AI policy playbook — the detailed version of this preliminary framework — is scheduled for release this month, June 2026, covering a system of 1.1 million students. ‍

New York City Public Schools is the largest school district in the country, which means that whatever it decides will have an outsized impact on how other districts approach the same questions. The rollout follows similar efforts already underway in other large US districts, including Chicago, Denver, and Charlotte-Mecklenburg, as well as statewide policies in places like Ohio. What NYC publishes this June will likely become a template that ripples through American public education far beyond its own five boroughs. ‍

The traffic-light system — and what it actually permits

The guidance uses a traffic-light rubric: red for never, no exceptions; yellow for proceed carefully with human oversight; green for encouraged. The red list is specific: AI cannot make decisions about placement, discipline, graduation, or program access. IEPs and 504 plans stay with qualified humans. Grading stays with the teacher of record. OpenAI

Under the guidance, teachers may use AI for brainstorming, organising, drafting communications, and lesson planning, but AI cannot be used to assign grades, make disciplinary decisions, or collect biometric and behavioural data without strict oversight. The guidance encourages teachers to use AI to explore lesson ideas, approaches, and unit planning, and for teacher training, scheduling, and resource planning. It also suggests AI can support better educating students with disabilities or who do not speak English at home — though translations and disability accommodations must be reviewed by qualified staff. ‍

Critically, the guidance is explicit about the limits of AI's role: "AI supports — it never replaces — educator decision-making." In classrooms, the policy allows students to use AI for research and exploration, but requires educators to verify outputs for accuracy and bias. Teachers must also instruct students to question AI responses rather than accept them at face value. ‍

That last sentence is, in many ways, the heart of the entire framework. NYC is not asking teachers to police AI use from the outside. It is asking them to actively teach students the habit of questioning what AI produces — to build scepticism and critical evaluation into the normal use of these tools, rather than treating AI outputs as settled answers.

The honest gap NYC has admitted to

What makes this guidance genuinely unusual, compared to most institutional AI policy documents, is its candour about what it cannot yet do.

The review system, known as the Enterprise Request Management Application, requires vendors to disclose how they handle student data, confirm compliance with federal and state privacy laws, and agree not to train AI models on student information. Tools that fail the process cannot be used in schools. But ERMA currently evaluates data privacy. It does not yet evaluate algorithmic bias, equity impact, or instructional effectiveness — capacity that NYC Public Schools says it is actively building, with the expanded review reflected in the upcoming June Playbook. ‍

The vetting process does not yet include guidelines on how to review certain aspects of AI products, such as algorithmic bias or instructional effectiveness. Those are expected to be included in the final June version of the playbook.

This is a significant admission from an institution governing 1.1 million students. NYC has built robust privacy infrastructure — verifying that an AI tool will not leak a child's data or train on it without consent — but it has openly acknowledged that it does not yet have a working method for answering a much harder question: does this AI tool work fairly for every child, regardless of their background, language, or learning profile? That capability is still being built. The honesty of saying so, rather than claiming a review process is comprehensive when it is not, is itself a notable choice in how an institution communicates about AI governance.

A Central AI Task Force, Data Privacy Working Group, and AI Advisory Council will govern ongoing implementation — and the Central AI Task Force includes 76 members from across the system's divisions. Seventy-six people, working through the question of what AI should and should not be allowed to do with 1.1 million children's education. That scale of institutional attention is itself a signal of how seriously this question is now being treated. ‍

The tension nobody has fully resolved

Parents packed a New York City Panel for Educational Policy meeting demanding the DOE pause all AI deployments in schools while the city finalises its governance framework. Critics argued that rolling out AI tools ahead of the DOE's own June 2026 playbook deadline puts students at risk, and that the preliminary guidance lacks enforceable safeguards and clear parental opt-out rights.

This is the structural tension that every school system attempting to govern AI will eventually face, and NYC's experience makes it visible in a way few other districts have been forced to confront publicly. AI tools are already in classrooms — students are already using them, whether or not a school has approved them, whether or not a policy exists to govern them. As one member of the Education Department's Data Privacy Working Group put it: "Just like TikTok was unregulated until school networks blocked it, so are these free AI products." Waiting for a perfect governance framework before acting means students continue using ungoverned tools in the meantime. Acting before the framework is complete means deploying tools under rules that are openly acknowledged to be incomplete. OpenAI

There is no clean resolution to that tension. NYC chose to move forward with what it has, while being transparent about what remains unbuilt. The moratorium demand from parents illustrates the political risk of the "deploy first, govern later" approach that many large urban districts have followed. Whether that approach proves wise will only be clear in retrospect — but the debate itself is one every school community, anywhere in the world, will eventually need to have.

Why this matters for AI Ready School and every school we work with

NYC's guidance, read carefully, is not really a story about New York. It is the most detailed public articulation yet of a question AI Ready School has been built around from the beginning: how does a school use AI to genuinely help students and teachers, without quietly outsourcing the judgment, care, and human relationship that education actually depends on?

"AI represents a stress test of our ability to stay aligned to our public mission. The children who need the most support are already in classrooms where AI is part of the tools they use. Governing well is the condition under which equity is possible." That sentence, from NYC's own guidance, could sit comfortably inside AI Ready School's own philosophy. The children who most need thoughtful, well-governed AI are not hypothetical. They are the same students our use cases describe — the silent struggler, the anxious achiever, the differently-wired learner — and an AI tool that has not been evaluated for bias or equity impact is a tool that may quietly fail exactly the students who can least afford it to fail.

This is precisely why Morpheus is built the way it is: as a tool that hands teachers more time and better insight, while leaving the decisions that matter — grading, intervention, how to respond to a struggling child — firmly with the human in the room. "AI supports—it never replaces—educator decision-making" is not a constraint AI Ready School works around. It is a principle our products are designed to embody. Cypher is built to make students question, evaluate, and think critically about what an AI tells them — the same habit NYC's guidance asks teachers to actively instil in their classrooms. Teachers must instruct students to question AI responses rather than accept them at face value — that sentence describes, almost exactly, the design philosophy behind Cypher's refusal to simply hand students answers. ‍

The bias and equity review NYC has not yet built is also a reminder of why a thoughtful, evidence-based approach to AI in education cannot be assumed — it has to be actively constructed, tested, and continuously examined. As AI Ready School works with schools across India and internationally, the same standard applies: does this tool work fairly for every learner, in every context, regardless of language, background, or learning profile? That question does not have a final answer. It requires ongoing, honest evaluation — exactly the kind NYC has committed to building, and exactly the kind every serious AI-in-education provider should hold itself to.

The sentence worth remembering

"The question is not whether AI belongs in schools. The question is whether we will collectively build a system that governs AI to serve every student and every stakeholder."

That is the largest school district in America, writing for 1.1 million children. It is also, in essence, the question every school in the world is now being asked to answer for itself.

‍