The School AI Policy Kit

A free, vendor-neutral template for writing your school's own AI policy. Built from conversations with more than a hundred educators.

Prefer to fill it in yourself? Download the editable .docx.

How to use this kit

Read through the sections below to see the decisions a real AI policy has to make. Then download the editable version, fill in the blanks with your school's own answers, and change anything that does not fit. This template is built to serve your school and your students first, and to be useful even if you never adopt Deskpad Learn or any other tool. Everything here is free, and students are never charged.

The template

Eight decisions every AI policy has to make. Read the thinking here, then fill in the blanks in the editable version.

1. 1

   Your stance

   Start by writing down where your school stands, in a sentence or two, so every decision that follows has a north star. Steer clear of both 'we ban AI' and 'anything goes.' Name the principle instead: AI is welcome when it strengthens a student's thinking and stays visible, and discouraged when it replaces the work the student is there to do.

   Decision to record

   Your school's one-sentence position on AI.

   • In student work, our school believes AI should be __________.
   • AI use is encouraged when __________.
   • AI use is not allowed when __________.
2. 2

   Where AI is allowed, and how students show their work

   This is the heart of the policy. Spell out which uses are welcome (brainstorming, feedback, explaining a hard concept), which are off-limits (generating a draft and turning it in as your own), and how students make their process visible. Transparency is what makes the rest workable. When the thinking is shown, AI use stops being something to hide.

   Decision to record

   Your allowed-use list, your not-allowed list, and how students disclose AI use.

   • Allowed uses: __________.
   • Not allowed: __________.
   • Students disclose AI use by __________ (for example, a short note on the assignment, or a visible revision history).
3. 3

   Keeping thinking visible

   The real risk is not AI itself, it is cognitive offloading: students handing over the part of the brain that thinks critically. The research on over-reliance is real. The mitigation is to treat AI as a thought partner rather than a ghostwriter, and to keep some visibility into how work gets made, the way schools always have. When AI is allowed thoughtfully, students start to learn for themselves where it helps them and where it hurts them.

   Decision to record

   How your school keeps the student's own thinking visible in graded work.

   • We keep student thinking visible by __________.
   • When we suspect work was offloaded, we __________.
4. 4

   Energy and environmental cost

   Be honest here, because the cost is real. Training and running large models uses meaningful energy and water. Two things keep it in proportion. Most of the footprint is in training, a cost paid once and then spread across billions of uses, so a single student's question is a tiny fraction of it. And efficiency is improving, with smaller and local models doing more on ordinary hardware. The honest path is not to single out AI for a ban while every other energy-hungry technology in the building goes untouched. It is to adopt it deliberately and prefer the most efficient tool that meets the need.

   Decision to record

   Your school's stance on environmental cost, and any preference for efficient or local tools.

   • Our stance on the environmental cost of AI: __________.
   • Where possible, we prefer tools that __________.
5. 5

   Student data and privacy

   Decide what student data is allowed to leave the building, and through which tools. Today most AI messages go to a cloud provider, and those providers are held to strict data-privacy standards with heavy penalties for misusing data, especially for minors. The field is also moving toward local models that run on a school's or a student's own hardware, which removes the concern entirely. Until then, approve specific tools deliberately and read what they actually do with student data.

   Decision to record

   Approved tools, and what student data may and may not be shared.

   • Approved AI tools: __________.
   • Student data that may never be shared with an AI tool: __________.
   • We review a tool's privacy terms before approving it: yes / no.
6. 6

   Accuracy and hallucination

   AI can state false things confidently. The most effective safeguard is not waiting for a perfect model, it is changing how the model is used: ground it in trusted source material like the syllabus or the assignment, let it say 'I don't know,' and teach students to verify. Tools built this way, anchored to vetted material rather than answering from memory, cut error rates dramatically.

   Decision to record

   How your school handles AI accuracy and teaches verification.

   • We expect AI tools used in class to be grounded in __________.
   • We teach students to verify AI output by __________.
7. 7

   Rolling it out

   A policy nobody understands is not a policy. Plan how you will explain this to students, parents, and staff, in plain language, before it takes effect. Say what changed, why, and what is expected, and invite questions.

   Decision to record

   Who you tell, when, and how.

   • We will share this policy with students by __________.
   • We will share it with families by __________.
   • Staff training and support: __________.
8. 8

   Reviewing it

   AI moves fast, so your policy should be a living document. Pick a cadence to revisit it, at least once a year, and name who owns the update so it does not quietly go stale.

   Decision to record

   Your review cadence and the person who owns it.

   • We review this policy every __________.
   • The person responsible for keeping it current is __________.