Part III of the AI in school series.

Rhode Island says the quiet part out loud. “Artificial intelligence is not the future for our schools – it’s the present,” Commissioner Angélica Infante-Green announces, pairing that posture with an explicit promise of “implementation support and professional learning opportunities” plus an AI advisory group to keep the work moving (Rhode Island Department of Education, 2025).

Massachusetts stakes out a roadmap teachers can plan around. The Department’s Office of Educational Technology lays down a multiyear sequence that includes resource creation in 2025, and then a dedicated year for workshops, trainings, tool recommendations, and technical assistance across 2025–26, with policy integration and educator-prep alignment in 2026–27 (Massachusetts Department of Elementary and Secondary Education, 2025).

Maine leans into the same idea but makes the training tactile. The DOE releases an “AI Guidance Toolkit” that “pairs guidance with learning,” promising practical scenarios and ongoing professional learning so districts aren’t left hunting for examples (Maine Department of Education, 2025).

Zoom in and the tone is clear: don’t ban, steer. The Boston Globe captures that moment in New England - teachers already using AI, administrators asking for a map, and states signaling green lights with guardrails (Machado, 2025).

New Hampshire, as usual, moves through pilots. The state partners with Khan Academy, giving grades 5–12 free access to Khanmigo, the AI teaching assistant, and keeps extending that access while promoting PD around it (New Hampshire Department of Education, 2024; 2025). “The AI is here, and it’s not going away,” a Granite State principal says, cutting through the noise (DeWitt, 2025).

New York City, the bellwether, reverses its early ban and starts growing internal professional learning streams through district innovation teams and partners like the New York Academy of Sciences. The posture shifts from fear to fluency: “we could run, but we could not hide from this technology,” Chancellor David Banks says - nudging the system toward teacher-led, human-in-the-loop use (Varanasi, 2023; St. John’s University, 2024; NYC Digital Learning & Innovation, 2025).

Across all five Northeastern states here you can feel the same pattern: voluntary statewide guidance, pilots to de-risk, and a heavy emphasis on PD. The federal Dear Colleague Letter in July 2025 strengthens that posture by clarifying that AI-related purchases and training can be supported with existing grant programs - a small sentence that changes district budgets (U.S. Department of Education, 2025).

What teachers actually have in their hands

Teachers don’t live in memos; they live in minutes. Survey after survey shows the gap between vision and training. In fall 2024, Education Week’s nationally representative survey found that 58 percent of teachers had received no PD on generative AI; only 6 percent reported ongoing training (Education Week Research Center, 2024). Another scan showed just 48 percent of districts had trained teachers by fall 2024 - up sharply from the year prior, but still a coin flip (Diliberti et al., 2025).

Those numbers land like a weight in a staff meeting. Because if AI is now expected in lesson planning, formative checks, parent comms - no training means uneven adoption. And uneven adoption means the risk shifts to the individual teacher’s judgment. Or their courage.

You can see states trying to close that gap with structures teachers recognize. Massachusetts promises workshops and coaching windows that map to the school year. Maine packages scenarios teachers can lift right into practice. New Hampshire ties PD to actual tools teachers will use tomorrow, not someday. Rhode Island promises a standing advisory group so the guidance doesn’t calcify (Massachusetts Department of Elementary and Secondary Education, 2025; Maine Department of Education, 2025; New Hampshire Department of Education, 2024; Rhode Island Department of Education, 2025).

Private and charter schools move faster under the same sky. In New York City, the Digital Learning & Innovation team opens citywide PD while networks and independents layer their own sprints - STEM-aligned workshops through the New York Academy of Sciences, campus-level coaching, and faster procurement cycles (NYC DLI, 2025; New York Academy of Sciences, 2025). In the Northeast, autonomy often means teachers see the training before the regulation. That speed is a gift - until it isn’t.

One more wrinkle: national unions and vendors are stepping into the PD vacuum. The AFT’s National Academy for AI Instruction, seeded by big-tech dollars, aims to train hundreds of thousands of teachers this decade. Teachers will care less about who funded it than whether it is practical, human-scale, and aligned to their state’s guardrails (Merod, 2025; Time, 2025).

A different bet on coherence

Now place all of that against China’s national playbook. The Ministry of Education issues two complementary guidelines in May 2025 - one for “general AI education” across K-12 and one for the “use of generative AI.” They spell out stage-by-stage expectations, restrict unsupervised generative AI use by young students, and make an explicit rule teachers in the Northeast can only infer: AI may not substitute for a teacher’s core role (Ministry of Education of the PRC, 2025).

Cities then translate policy into timetables. Beijing’s schools will deliver no fewer than eight hours of AI instruction per year, folding evaluation results into students’ comprehensive assessments. Teachers are expected to use AI for feedback and diagnostics - but inside a frame the city sets (Global Times, 2025; Business Insider, 2025).

Hangzhou goes further - mandating AI across primary and secondary schools with at least ten class hours annually, and explicitly naming teacher training in the municipal rollout (South China Morning Post, 2025).

Watch the difference through a teacher’s eyes. In China, the work is directive: minimum hours, staged competencies, named prohibitions, and a through-line from national intent to local delivery. PD is not an optional layer - it is embedded in the architecture. Reuters describes the broader reform as using AI to elevate independent thinking, problem-solving, and collaboration, but under rules that secure data and preserve teacher authority (Reuters, 2025; eWeek, 2025).

In the Northeast, the work is accretive. State guidance is voluntary. Districts and networks test. Teachers piece together a practice from pilots, tool menus, and scattered PD days. Coherence lives inside the school more than the system.

Mass upskilling vs. grassroots PD

Here’s the hinge. China’s model is mass upskilling by decree - teacher competencies specified, student exposure mandated, curricular hours counted. It’s fast to coverage, slower to dissent. The benefit for teachers is predictability: what to teach, how often, what not to do. The cost is the risk of brittleness when the technology shifts.

The Northeast’s model is grassroots PD under permissive guidance - teachers learn by doing, states scaffold with resources, unions and nonprofits fill in gaps. It’s slow to coverage, faster to dissent. The benefit for teachers is agency. The cost is inequity between classrooms and zip codes.

Teachers feel both truths at once. “We have to buckle down and understand it, so that we can help kids be prepared for it,” a New Hampshire learning leader says - her sentence almost a mirror of Rhode Island’s official line about the present tense of AI (DeWitt, 2025; Rhode Island Department of Education, 2025).

But teachers also name the training vacuum. Education Week’s data lands hard: the majority without PD, a small fraction with ongoing coaching, and quality ratings all over the map (Education Week Research Center, 2024). And parents, boards, and students are already asking for clarity. That’s the job, teaching under ambiguity, while building the policy that reduces it.

What to give teachers tomorrow morning

If you’re in a Northeastern classroom, the governance story is only helpful if it becomes time. A usable day looks like this: a vetted tool that respects student privacy; a state-aligned protocol for how you disclose, verify, and archive AI-assisted materials; coaching cycles tied to your actual content; and a shared language that says when to reach for AI, when not to. That’s human-in-the-loop as practice, not slogan.

Massachusetts’ timetable, Maine’s toolkit, New Hampshire’s pilot-plus-PD, Rhode Island’s advisory spine, New York’s networked workshops, those are the moves that turn policy into prep.

China’s bet, if you’re looking across the ocean, is less about magic and more about sequence. Eight hours. Ten hours. Spiral curriculum. No substitution of teacher judgment. Two playbooks, then. Different speeds, different forms of certainty. The same classroom stakes.

Teachers will make this real either way.

References