China & the US Northeast are setting AI’s Ground Rules for K-12

Part I of the AI in schools series.

The emails from principals landed all summer: “What’s our AI policy?” Teachers read them between PD slides and last-minute room prep, thumb hovering over a dozen tabbed “guidance” PDFs. Across New York, Massachusetts, New Hampshire, Maine, and Rhode Island, the message is similar yet not the same. Use AI to save time. Use it to individualize. Vet the tools. Protect kids. But when you zoom out, the pattern isn’t one pattern at all—it’s a mosaic. And when you look across the Pacific, China is painting a very different picture: a mandate.

New York’s pivot: from “knee-jerk fear” to cautious embrace

New York City reversed course early. After restricting ChatGPT in January 2023, Chancellor David C. Banks went public that May with a different stance, acknowledging the rollout “sparked over 1,000 news articles” and that the initial response “signaled fear and risk.” He wrote plainly: “The knee-jerk fear and risk overlooked the potential of generative AI to support students and teachers.” (Banks, 2023)

Banks promised support and exemplars—“a repository and community to share findings”—and pointed to classroom moves already underway, like teacher-guided analysis of AI outputs in Queens. The point wasn’t to drop guardrails; it was to move them—away from blanket bans and toward teacher-led exploration. (Banks, 2023)

For rank-and-file educators, that shift matters. A ban gives you certainty but no tools. A permission structure backed by PD, templates, and “what good looks like” lets you try AI without feeling like the pilot program yourself.

Massachusetts sets a multi-year runway

Massachusetts is not asking every school to invent its own playbook in isolation. The Department of Elementary and Secondary Education (DESE) convened a K-12 AI Task Force in 2024 and published a multi-year roadmap in 2025: resource creation, professional learning, and—later—embedding AI into curriculum frameworks and educator preparation. Translation for teachers: the state is building upstream capacity so downstream classrooms aren’t improvising policy between periods. (Massachusetts Department of Elementary and Secondary Education, 2025)

The DESE page spells out timing—resources completed in Summer 2025; workshops, tool recommendations, and technical assistance through 2025-26; policy considerations and integration into frameworks in 2026-27. That pacing buys districts time to align PD, procurement, and parent communication without leaving teachers in the lurch. (Massachusetts Department of Elementary and Secondary Education, 2025)

Learn more about AI

Stay updated with new articles about AI and prompting!

Subscribe

Email sent! Check your inbox to complete your signup.

No spam. Unsubscribe anytime.

Rhode Island’s message: adopt carefully

Rhode Island took a direct tone this August, releasing statewide guidance framed as a “first step” toward “responsible and effective” AI use. Commissioner Angélica Infante-Green said it flatly: “Artificial intelligence is not the future for our schools—it’s the present.” The guidance urges AI literacy, vendor vetting, and district-level policies, without pretending the risks disappear. And it names the current reality most teachers recognize: students are already using AI. RIDE’s survey found roughly 20% of students report using tools like ChatGPT or Grammarly, while only about 6% of educators do so regularly. (Rhode Island Department of Education, 2025)

Reporting in The Boston Globe captured the teacher’s dilemma well: some districts are testing educator-facing tools such as Magic School AI after signing student-data agreements, while keeping general-purpose chatbots blocked over privacy concerns. It’s adoption with brakes—use the vetted wrench, keep the workshop door locked. (Machado, 2025)

Maine makes guidance interactive

Maine’s Department of Education launched an interactive AI Guidance Toolkit in February 2025, pairing policy framing with practical scenarios and ongoing learning sessions. It’s not a memo you file; it’s a living resource that walks through “what now?” in classroom-ready form. For teachers who need examples more than slogans, that style of guidance meets the moment. (Maine Department of Education, 2025)

New Hampshire: norms before noise

New Hampshire is leaning into norms and supports. The state partnered with Khan Academy to provide Khanmigo at no cost to schools and highlighted instructor-led use with privacy front-and-center. “AI is a tool. It doesn’t replace educators,” said newly confirmed Commissioner Caitlin Davis this summer. Local groups produced a 77-page guidance document focused on task forces, PD plans, and simple heuristics teachers can use—be purposeful, be critical, be ethical. You can feel the emphasis: don’t chase features; align to instruction first. (DeWitt, 2025)

Federal green light, human hands on the wheel

On July 22, 2025, the U.S. Department of Education sent a Dear Colleague letter that landed like a clarifying note. Districts can use existing federal grant funds to buy or build AI tools, support tutoring, craft instructional materials, and train educators—“without replacing the critical role they play.” It’s an important sentence. For teachers, that line legitimizes experimentation and gives CFOs and superintendents the compliance cover to fund it. (U.S. Department of Education, 2025a, 2025b)

Private and charter schools: faster pilots, same duty of care

Because procurement is simpler, private and charter schools often pilot faster—but they’re not exempt from the same obligations teachers worry about: FERPA, vendor data practices, transparency with families. In Rhode Island, Narragansett teachers tested Magic School AI only after the vendor signed a student-data privacy agreement; open general chatbots remain blocked on district networks. That pattern—enterprise contract first, classroom use second—is increasingly common across charters and independents in the Northeast. (Machado, 2025)

China’s approach: clear lines, compulsory hours

Meanwhile, Chinese authorities are drawing bold lines teachers can see from the hallway. Beijing mandated “at least eight class hours” of AI instruction yearly for all primary and secondary students starting fall 2025, delivered as standalone or integrated courses. The plan sequences by age: experiential awareness in primary, cognitive application in junior high, and practical innovation in high school. This isn’t a pilot—it’s a citywide requirement. (Xinhua, 2025)

Hangzhou went further, announcing compulsory AI classes with a minimum of 10 hours annually and flexibility in delivery—one concentrated week or integrated units—with parallel training for teachers. That specificity—time on task, scope, and methods—shifts the burden off individual educators. The hours aren’t a target; they’re a floor. (Nulimaimaiti, 2025)

At the national level, China’s Ministry of Education issued guidelines in May 2025 that both promote AI education and restrict use by age. Primary students are barred from independently using open-ended generative tools; teachers are explicitly told AI must “complement but not replace” human-led teaching. It’s prescriptive, and it’s uniform. (Zou, 2025)

What the rules feel like in a classroom

Policy tone shapes teacher posture. In Beijing or Hangzhou, the ask is straightforward: teach AI for a set number of hours, with guardrails spelled out by grade and a clear rationale of national talent development. Teachers still have all the work—planning, differentiation, formative checks—but they don’t spend August arguing if AI belongs in class, or whether a principal will back them if parents object. The system has decided.

In the Northeast U.S., teachers are being asked to decide with their systems. Massachusetts is building a runway. Rhode Island is pressing “go,” with cautions. Maine is meeting teachers where they plan. New York City is encouraging exploration after acknowledging the initial overcorrection. New Hampshire is normalizing instructor-led use while codifying common-sense rules. And Washington has told districts the money can follow—if the work stays responsible. (Rhode Island Department of Education, 2025)

Both approaches can work. The question is where the cognitive load sits. In the mandate model, teachers inherit a schedule and a syllabus spine; their craft lives in enactment. In the mosaic, teachers inherit discretion; their craft lives in judgment. Some will prefer one burden to the other. All deserve clarity.

Teachers don’t need another shiny tool

They need time back for feedback, small-group instruction, and building students’ stamina for thinking. That’s the promise many educators see in educator-facing AI—drafting lesson shells, differentiating texts, generating practice with answer rationales. But those gains evaporate if every district—or every school—is reinventing policy from scratch. The Northeast states moving fastest are doing the unglamorous work: procurement guardrails, PD pathways, privacy vetting, and exemplar libraries teachers can copy, adapt, and trust. (Massachusetts Department of Elementary and Secondary Education, 2025)

Meanwhile, China’s clarity will generate different questions teachers everywhere recognize: twelve months later, did mandated hours translate into real student understanding? Did scripted scope and sequence leave room for teacher judgment? Did top-down guardrails evolve as classroom realities changed? Uniformity reduces uncertainty; it can also reduce the oxygen teachers need to iterate.

From a teacher’s standpoint, the difference between mandate and mosaic isn’t theoretical. It shows up in your planning period. It shows up in how your principal answers a parent email. It shows up in whether your district will buy the tool you actually use—or block it while urging “innovation.” In the Northeast, the center of gravity is moving toward support over bans and toward teacher-led, instructor-first use with privacy guardrails. In China, authorities are pushing consistency, time on task, and staged exposure from grade one onward.

Classrooms will make or break both models. That’s where the ethics become habits, where the hype meets pacing guides, and where kids decide whether AI is a shortcut or a scaffold. Whatever the policy, the work is the same: keep the human judgment at the center, and make the rules serve learning—not the other way around.