Creating a School AI Policy: A Practical Guide for UK Leaders

AI in education is no longer optional. Senior leaders need a clear, written policy that protects pupils, supports staff, satisfies the DfE's expectations and aligns with UK GDPR. This guide gives you a practical framework — grounded in doctoral research at the University of Oxford and how UK schools are actually using AI today.

Why your school needs an AI policy now

The DfE's Generative AI in education position paper makes clear that schools are responsible for how AI is used on their premises and in their teachers' workflows. Ofsted will increasingly expect leaders to demonstrate that AI use is intentional, safe and tied to the curriculum — not improvised. A written policy is the artefact that proves it.

UNESCO's Guidance for generative AI in education and research and the NEA's AI policy frameworks both converge on the same core principles: human oversight, data minimisation, transparency with families, and equity of access. Your policy should make those principles concrete for your context.

The seven sections every school AI policy should contain

1. Purpose and scope. Who the policy covers (teachers, support staff, pupils, governors, third-party tools) and what counts as "AI use" at your school.
2. Approved tools and procurement. A short list of AI tools the school has assessed, with the SLT-approved use case for each. New tools require a DPIA before adoption.
3. Data protection and UK GDPR. No pupil personal data, SEND information or safeguarding records may be entered into general-purpose AI tools. Where AI is used with pupil data, contracts must include UK GDPR-compliant data processing terms.
4. Staff use. Encouraged uses (planning, marking support, differentiation, communication drafts), required uses (human review of every output), and prohibited uses (entering pupil names, safeguarding details, or third-party copyrighted material).
5. Pupil use. Age-appropriate guidance, academic integrity expectations for coursework, and how AI literacy is taught across the curriculum.
6. Bias, accessibility and SEND. Recognition that AI outputs can reinforce bias; commitment to checking outputs against your school's SEND register and inclusion principles.
7. Review, training and accountability. Named SLT lead, annual review cycle, mandatory staff training, and a clear route for raising concerns.

Ethical governance: who decides?

The single most important governance decision is naming an AI lead on the SLT — typically a deputy head or assistant head with a digital remit. That person owns the approved-tools list, signs off DPIAs, and reports annually to governors. Without a named owner, AI policy becomes everyone's problem and therefore nobody's.

Governing boards should add AI to the standing agenda for the curriculum, safeguarding and data committees, not treat it as a one-off update.

Data privacy: the non-negotiables

• No identifiable pupil data — names, photos, SEND categories, safeguarding flags, contextual data — in general-purpose AI tools (ChatGPT, Gemini, Claude consumer accounts).
• School-approved platforms must offer a UK or EEA data residency option and a contractual commitment not to use school data for model training.
• Any AI tool that ingests pupil work must be covered by a Data Processing Agreement (DPA) signed by the school.
• Anonymisation (e.g. "Pupil A", "Pupil B") is the default working practice for staff using AI for planning and analysis.

Implementation: a 90-day rollout

Days 1–30 — Audit. Survey staff about what AI tools they already use. Identify shadow IT. Inventory what pupil data those tools have touched.

Days 31–60 — Draft and consult. Write the policy using the seven sections above. Share with governors, union reps, the DPO and a sample of teachers. Iterate.

Days 61–90 — Train and launch. All-staff INSET on the policy, with worked examples. Publish a parent-facing one-pager. Add the policy to your safeguarding induction.

Templates and tools that speed this up

Writing a policy from scratch is slow. Most leaders we work with use the Teachers.guide Policy Draft Generator to produce a first draft in minutes — grounded in UK GDPR, the DfE position paper and your school's context — then edit it with SLT colleagues. The generator outputs a policy that already includes the seven sections, a DPIA prompt list, and a parent-facing summary.

Teachers.guide is itself built around these principles: school-controlled, SEND-aware, UK-curriculum-aligned, and covered by a school-level DPA. The platform is designed so that adopting it makes your AI policy easier to enforce, not harder.

What to do next

1. Name your SLT AI lead this term.
2. Audit current staff AI use — informally, without judgement.
3. Draft your policy using the seven-section framework, or generate a first draft with the Policy Draft Generator.
4. Take it to governors before publishing to staff.

A good AI policy doesn't slow your school down. It removes ambiguity, gives staff permission to innovate within clear limits, and reassures families that pupil data is safe. That is leadership.

Where policy meets platform

A policy is only as good as the tools staff actually use. Two guides cover the surfaces most likely to sit alongside your policy:

• School Educational Intelligence — the whole-school layer that turns policy into consistent practice.
• AI in Google Docs for teachers — how AI shows up inside the tool most staff already write in, and what that means for acceptable-use rules.