School AI Policy Template: Copy, Adapt, Adopt This Semester

What should a school AI policy cover?

A school AI policy defines who may use artificial intelligence tools, for what purposes, and under what conditions. It covers permitted staff uses, student use by grade band, data privacy, disclosure of AI-assisted work, academic integrity, prohibited uses, incident handling, and a schedule for review. This page is that policy template, ready to copy and adapt.

That's the whole job. Everything below exists to help you produce that document — an actual, adoptable one — without a committee spending a semester on it. The complete template sits a few sections down, ready to paste into a doc and start marking up today, followed by guidance on what to change for your context, because the adaptation is where the real decisions live.

The frame throughout is the person who usually inherits this task. Picture a middle school principal in a district where the superintendent has said "we should probably have something on AI" and nothing since: seventh-grade teachers quietly drafting materials with chatbots, an eighth grader who turned in an essay two teachers suspect wasn't hers, a parent email asking whether the school "allows ChatGPT," and no document to point any of them to. Writing the policy lands on whoever raises their hand. This page is for that person — and for the district-office equivalent staring at the same blank page with more schools attached.

One scope note before the template: this article covers the leader's side — writing the rules. If you're a teacher wondering what you're allowed to do before your school has a policy, that question is answered in our AI for teachers guide, including what the survey data says about how few schools have provided guidance.

A policy is no longer optional in a growing list of states

For two years, school AI policy was a should. It is becoming a must, and the shift is recent enough that many leaders haven't registered it.

Tennessee moved first: a 2024 law requires every local school board and public charter school to adopt a policy on AI use "by students, teachers, and staff for instructional and assignment purposes," implemented no later than the 2024–25 school year, with annual compliance reports to the state that must include the adopted policy itself. Ohio followed in its budget bill: every district, community school, and STEM school had to adopt an AI policy by July 1, 2026, and the state's education department published a model policy districts may adopt outright or customize. And in the 2026 legislative sessions, Idaho, Maryland, Oklahoma, and Virginia joined the list of states requiring districts to adopt AI policies — Oklahoma's version goes further than most, prohibiting AI from being the primary basis for grading, discipline, and other high-stakes decisions and requiring annual disclosure of AI tools to families with opt-out rights.

Even where no mandate exists, the reference material has piled up. By AI for Education's state-guidance tracker, 34 states plus Puerto Rico had published official K–12 AI guidance as of the tracker's October 2025 update, and ExcelinEd counted eight states with enacted AI-in-education laws as of May 2026, with active bills in more than twenty others. The federal government weighed in early: the U.S. Department of Education's May 2023 report, Artificial Intelligence and the Future of Teaching and Learning, made "Emphasize Humans in the Loop" its first recommendation — the report takes the stance that constituents "want AI that supports teachers and rejects AI visions that replace teachers" — and stated plainly that educational leaders must create policies to govern how AI is used in education.

Two vocabulary points from that landscape are worth stealing before you draft. First, the TeachAI toolkit — produced by Code.org, ETS, ISTE, Khan Academy, and the World Economic Forum with input from over a hundred organizations — distinguishes guidance (flexible, non-binding, easy to revise) from policy (formally adopted, longer-lasting, enforceable). Most of what states have published is guidance. What your board votes on is policy. Keeping the two layers separate in your own drafting is what lets you update the fast-moving parts without a board meeting; more on that in the adaptation section.

Second, you are not the first district to do this, so don't draft like it. North Carolina's education department published a full implementation guidebook in January 2024 — notably recommending responsible use over bans. Washington's OSPI released human-centered guidance the same month, built around an H→AI→H sequence: start with human inquiry, let AI produce, close with human review. Georgia's Leveraging AI in the K-12 Setting (January 2025) draws the line most policies need drawn: it sorts uses into high-stakes (writing IEP goals, evaluating educators, grading subjective work, predicting discipline) and non-high-stakes (drafting lessons, building rubrics, leveling texts) and tells districts to keep AI out of the first column. If your state has its own document, read it before the template below — where a state document and this template disagree, your state wins.

The template

Copy everything in this section into a working document. Bracketed [ADAPT] items are decisions your district must make; everything else is written to be adoptable as-is. It's deliberately tool-agnostic — it regulates behaviors, not brand names — which is most of why it won't be stale in a year.

[DISTRICT / SCHOOL NAME] Artificial Intelligence Acceptable Use Policy

Adopted: [DATE] · Next scheduled review: [DATE — see Section 11] · Policy owner: [ADAPT: title, e.g., Assistant Superintendent for Teaching and Learning]

1. Purpose and scope. This policy governs the use of artificial intelligence (AI) tools by all staff, students, contractors, and volunteers of [NAME] in the course of school business and schoolwork. It applies on and off campus and on district or personal devices, whenever the use relates to [NAME]'s instructional program or operations. It supplements — and does not replace — existing policies on technology acceptable use, academic integrity, student data privacy, and staff and student conduct. Where this policy conflicts with another, [ADAPT: name the controlling policy or the officer who resolves conflicts].

2. Definitions.

• AI tool: software that generates content (text, images, audio, video, or code) or produces predictions and recommendations in response to user input. This includes general-purpose chatbots, AI features embedded inside other approved software (writing assistants, slide generators, meeting summarizers), and education-specific AI platforms.
• Approved AI tool: an AI tool on the district's published approved list, which for any tool that touches student data requires a signed data privacy agreement with the district. The current list is maintained at [ADAPT: location or URL] by [ADAPT: role].
• Student data: personally identifiable information from education records as defined by FERPA, plus any information that could identify a student when combined with other available information.

3. Guiding principles.

• (a) A person is responsible for every decision. AI may inform or draft; it does not decide. Final judgments about grades, placement, discipline, evaluation, and consequential communication belong to the responsible staff member.
• (b) All AI use under this policy must be disclosable. If a staff member or student would be uncomfortable stating openly how AI was used in a piece of work, that use falls outside this policy.
• (c) Student data protection overrides convenience, without exception (Section 6).
• (d) This policy governs categories of use, not named products, so that it remains accurate as tools change.

4. Permitted staff uses. Staff may use approved AI tools — and, where no student data is involved, general-purpose AI tools — for:

• drafting and adapting instructional materials: lesson plans, worksheets, rubrics, question banks, and leveled versions of texts;
• drafting routine communications such as newsletters, reminders, and letters of a general nature, provided the sender reviews the draft and takes responsibility for its accuracy and tone;
• summarizing publicly available documents, brainstorming, and preparing meeting agendas and notes;
• [ADAPT: add or remove administrative uses — scheduling drafts, duty rosters, grant-narrative first drafts.]

The staff member using the tool remains accountable for the accuracy, fairness, and appropriateness of anything AI helped produce, exactly as if they had written it unaided.

5. Student use, by grade band. In grades 6–12, assignments carry one of three AI tiers, set and communicated by the teacher:

• Tier 0 — no AI. The work assesses skills the student must demonstrate unaided.
• Tier 1 — AI-assisted, disclosed. Students may use approved tools for defined support (brainstorming, feedback on a draft, practice questions) and must disclose that use (Section 7).
• Tier 2 — AI-integrated. Working with AI is part of the task itself, and the assignment says so.

Grade-band rules [ADAPT each band to your structure and your state's requirements]:

• Grades K–5: Students do not use generative or open-ended AI tools independently; district-approved adaptive learning platforms remain permitted under teacher supervision. Teachers may use approved tools with the whole class, on the teacher's own account and device. Students under 13 are not given accounts on consumer AI tools under any circumstances.
• Grades 6–8: Students may use district-approved, age-appropriate AI tools under teacher direction for Tier 1 and Tier 2 activities. When an assignment does not state a tier, the default is Tier 0.
• Grades 9–12: Students may use approved AI tools according to the tier stated on each assignment; teachers state the tier on every summative task. When no tier is stated, the default is [ADAPT: Tier 0 or Tier 1 — pick one and say it].

Vendor terms-of-service age minimums apply regardless of tier.

6. Data privacy and tool approval. No staff member or student may enter student data into an AI tool that is not covered by a signed data privacy agreement with the district. Prompts should describe tasks, not identifiable students. AI tools are vetted through [ADAPT: your existing edtech approval process and committee], using the same instrument as other software plus three AI-specific checks: the vendor states in writing that inputs are not used to train its models; data storage, retention, and deletion terms are documented; and the tool's conduct with minors' data complies with COPPA where students under 13 use it. Requests to approve a new tool go to [ADAPT: role] and receive a decision within [ADAPT: number] school days.

7. Disclosure and attribution. Students disclose AI use as the assignment's tier requires, in the format the teacher specifies — a one-line statement is sufficient unless the teacher asks for more: "I used [tool] to [what it did]; the final work is my own." Staff disclose AI assistance when asked directly, and any substantially AI-drafted communication to families or the public is reviewed and signed by the responsible staff member before it goes out. [ADAPT: decide whether routine internal staff materials require proactive disclosure; most districts do not require it.] No one covered by this policy may present required-disclosure AI work as entirely their own. Family notification: the district publishes annually the list of AI tools approved for student use and where families can read this policy. [ADAPT: where state law grants opt-out rights — Oklahoma's mandate does — name the opt-out procedure and who administers it.]

8. Academic integrity. Using AI outside the tier set for an assignment is an academic-integrity matter handled under [ADAPT: cite your existing academic integrity policy], not a technology infraction. Because AI-detection software produces false positives, a detector score alone is never sufficient evidence of a violation; determinations rely on [ADAPT: your process — draft history, version logs, a conversation with the student, comparison to prior work].

9. Prohibited uses (all users).

• Entering student data into an unapproved tool (Section 6).
• Using AI as the sole or final grader of student work, or to make or recommend placement, discipline, or special-education decisions.
• Adopting AI-generated content into an IEP or 504 plan without the case manager's review, revision, and adoption of it as their own, and the IEP team's decision. [ADAPT: some states' guidance treats IEP-goal drafting as a prohibited high-stakes use, and some statutes bar AI as the primary basis for such decisions — check yours before permitting any drafting at all.]
• Using AI to evaluate staff performance, narrow job applicants to a shortlist, or predict individual student outcomes for consequential decisions.
• Creating or sharing AI-generated images, audio, or video depicting any student or staff member, including so-called deepfakes.
• Using AI to harass, impersonate, or demean any person, or to circumvent district filters and age restrictions.

10. Incidents and questions. Suspected exposure of student data to an unapproved tool is reported to [ADAPT: title] within one school day and handled under the district's data-incident procedure. Deepfake or impersonation incidents go to [ADAPT: administrator role] under [ADAPT: code-of-conduct citation]; preserve the material, do not re-share it. Good-faith questions and self-reports about AI use are treated as professional learning, not misconduct. This policy fails if people hide their use, so the district commits to answering questions without penalty. [NAME] provides staff training on this policy at least [ADAPT: annually, commonly at back-to-school PD].

11. Review. This policy is reviewed every [ADAPT: 12 months recommended] by [ADAPT: committee], and sooner if state guidance or law changes, a new category of tool reaches students, or an incident reveals a gap. The approved-tool list (Section 2) and assignment-tier practices (Section 5) are administrative procedures that may be updated by [ADAPT: role] without reopening this policy. Version history is kept at [ADAPT: location].

Adapting it: the decisions the brackets are hiding

The template holds the structure; the brackets hold the politics. Three of them deserve more than a fill-in.

The board line. Decide deliberately which parts need board adoption and which stay administrative. A workable split: Sections 1–3, 6, and 8–11 — scope, principles, privacy, integrity, prohibitions, review — are stable enough to be board policy, and in mandate states (Tennessee's annual report to the state includes the adopted policy itself) that's the layer the law is asking for. The approved-tool list and the tier defaults are procedures; if changing them requires a board vote, they will simply never change, and the policy starts rotting the day it's adopted. Section 11's last sentence exists for exactly this reason.

Elementary versus secondary. An elementary school adopting this template should expect Sections 4 and 6 to do most of the work and Section 5 to be short: student use is teacher-mediated, the COPPA line is absolute, and the policy's audience is mostly staff and families. A middle or high school inverts that — Sections 5, 7, and 8 become the ones teachers quote, and the tier system needs to show up where students actually look, meaning the student handbook and the assignment sheet, not just the board minutes. Georgia's guidance makes the same point structurally: it tells districts to update grading, code-of-conduct, and acceptable-use policies together rather than bolting on a freestanding AI document nobody cross-references.

Who's in the room. Georgia's adoption process names the committee plainly: educators, parents, students, and IT staff, with a needs assessment before drafting. Add a special-education voice — two of the prohibited uses live in their world, and the line Section 9 draws between AI-drafted phrasing and the team's decision is worked through in our AI for special education guide — and have counsel read the final draft against your state's statute if you're in a mandate state. A middle school principal drafting solo can absolutely produce the first version from this template in an afternoon; what they shouldn't do is take that solo draft straight to adoption, because the committee pass is where the tier defaults get pressure-tested by people who will actually enforce them.

Where school AI policies go wrong

Reading real district policies alongside the state guidance, the same failure patterns keep appearing — worth checking your draft against each one before it goes anywhere near an agenda.

The all-ban policy. Blocking every AI tool reads as decisive and performs as abdication. The ban stops at the school network: students carry the same tools in their pockets, staff keep using them at home, and the district converts visible, teachable use into hidden use it can no longer see, guide, or build disclosure habits around. This is why North Carolina's guidebook recommends implementation over prohibition, and why the better-designed frameworks — Georgia's high-stakes line, Oklahoma's statute — ban specific uses rather than the technology. A policy whose only verb is "prohibited" also expires the moment the district adopts any AI-featured product, which, given what's now embedded in standard productivity suites, has usually already happened.

The students-only policy. Some drafts regulate what a fourteen-year-old may do with a chatbot and say nothing about the adults. That's backwards twice over: the highest-stakes risks in the building — student data pasted into unapproved tools, AI-drafted IEP content, automated grading of subjective work — are staff behaviors, and students notice asymmetry immediately. If your draft's Sections 4, 6, and 9 are thin while Section 5 runs two pages, rebalance before adopting.

The policy frozen in 2023. Plenty of districts wrote something in the first ChatGPT year and haven't touched it since. Those documents ban tools by name that have been renamed twice, never mention AI features inside software the district itself licenses, and predate every state document cited on this page. The fixes are structural, not stylistic: tool-agnostic definitions (Section 2), an approved list that lives outside the policy, and a review clause with named triggers (Section 11). If your current policy has no review date, treat that as the finding.

The policy with no procedure underneath. Adopting the document and stopping there produces a policy that answers no actual question a teacher has on a Tuesday. The template assumes a living layer below it — the tool list, the tier stated on each assignment, the named person who answers "can I use this?" within a set number of days. If those brackets got filled with vague nouns instead of names and numbers, the policy is a press release.

The two fears in the room

Every AI policy conversation carries two anxieties that the document itself has to answer, because they're what people are actually thinking about during the public-comment period.

The first is cheating. Parents and teachers will read your policy looking for one thing: does this school still know whether students did their own work? The honest answer is that a policy can't detect anything — but it can make expectations legible in a way that vague syllabus warnings never did. The tier system does that: "no AI on this essay" and "AI brainstorming allowed, disclosed" are enforceable sentences, where "use AI responsibly" is not. And the detector clause in Section 8 protects the school from the opposite failure — a false accusation built on a probability score is a faster way to lose a community's trust than any undetected chatbot essay.

The second fear is quieter and belongs to your staff: being judged for using AI at all. A teacher who drafts materials with a chatbot and hides it isn't hiding from the policy — there usually isn't one — but from colleagues. A policy that names permitted staff uses in writing does something a staff-meeting reassurance can't: it makes honest use cheaper than hidden use. That's also the case for writing Section 10's amnesty line and meaning it. The privacy rules in Section 6 only work if the teacher who realizes they pasted something they shouldn't have reports it within a day, and nobody reports into a policy that treats the report as a confession. (For what that data exposure actually means legally — FERPA's definitions, what a DPA commits a vendor to, COPPA's under-13 line — the reasoning is in our student data privacy guide; this policy's Section 6 is the enforcement surface for all of it.)

Adopt the template, fill the brackets with real names and numbers, and put a review date on it before anyone votes.