Keeping Children Safe in Education 2026 came into force on 1 September — and for the first time, it deals directly with AI. Here's what changed, what your school needs to do, and how to get it sorted in a single 30-minute call.

If you're a headteacher, DSL or governor at a UK primary school, the 2026 update to Keeping Children Safe in Education (KCSIE) is the one document you can't skim this term. The Department for Education has, for the first time, written generative AI directly into statutory safeguarding guidance — and that changes what "compliant" looks like for every school in England.

The uncomfortable truth: most primary schools currently have, at best, a holding AI policy — a page or two written in 2023 that says "staff should use AI responsibly." Under KCSIE 2026, that no longer cuts it.

This guide covers exactly what changed in KCSIE 2026, what your school needs to do about it, and how to get a complete, governor-ready set of AI documents without adding weeks to your workload.

What is KCSIE 2026?

Keeping Children Safe in Education is the DfE's statutory safeguarding guidance. Every school in England — primary, secondary, maintained, academy — must follow it, and every member of staff must read at least Part One. It's updated most Septembers, and the 2026 edition took effect on 1 September 2026.

Most years the changes are modest. This year, four of them relate directly to artificial intelligence — and together they mean your school's AI policy, staff training and parent communication all need reviewing this term.

What Changed in KCSIE 2026 on AI? The 4 Key Changes

1. Generative AI is now in statutory safeguarding guidance

KCSIE 2026 adds new paragraphs specifically on generative AI. Schools are now expected to understand the safety considerations and legal responsibilities that come with choosing to use generative AI tools — whether those tools are teacher-facing (lesson planning, report writing) or pupil-facing. The guidance points schools to the DfE's generative AI in education material and its product safety expectations, covering data protection and legal duties.

What this means for your primary school: "We haven't really decided on AI yet" is no longer a neutral position. If staff are using ChatGPT, Gemini or Copilot for planning or reports — and in most schools, some quietly are — the school is expected to have understood and managed the risks. (If that's happening in yours, start with our guide to whether teachers can use ChatGPT for reports safely.)

2. Deepfakes and AI-generated images are explicitly a safeguarding issue

The guidance now makes clear that images in a safeguarding context may be digitally altered or entirely AI-generated, including deepfakes. The old grey area — "but it isn't a real photo" — is closed. AI-generated imagery of children is treated as the safeguarding matter it is.

What this means for your primary school: Your child protection policy, DSL training and staff awareness all need to reflect AI-generated and manipulated imagery. Primary-age children are increasingly exposed to these tools through older siblings and social platforms.

3. AI chatbots are now treated as a contact risk

KCSIE's online safety framework categorises risk into content, contact, conduct and commerce. The 2026 update explicitly brings generative AI applications that simulate harmful online interaction into the contact risk category. In plain English: an AI companion chatbot is now viewed like a stranger online, not just content to filter.

What this means for your primary school: Your online safety curriculum and filtering approach need to account for AI chatbots and companion apps — not just websites. This also needs explaining to parents, because most of this exposure happens at home. (For the wider risk picture, see our piece on the safeguarding risks of AI companion chatbots.)

4. Filtering and monitoring must be reviewed every year — with AI in scope

Schools must now review the effectiveness of their filtering and monitoring at least once every academic year — and that review needs to consider AI tools, including how pupils might access or bypass restrictions to reach them.

What this means for your primary school: Someone — usually the DSL with IT support — needs to own an annual filtering and monitoring review, documented, with AI tools in scope. Governors should see it.

Does My School Need a Standalone AI Policy Under KCSIE 2026?

Here's the nuance most articles miss: KCSIE 2026 does not mandate a single document called "AI Policy." What it expects is that AI runs through your whole safeguarding approach — your child protection policy, staff code of conduct, online safety teaching, parent engagement, DSL role and filtering arrangements.

That's actually a higher bar than one document. A school can't meet it by adding a paragraph about ChatGPT to an old policy. But in practice, the cleanest way to evidence it — to governors, to parents, and to Ofsted — is a coherent set of documents:

  1. An AI policy — the school's position on AI use by staff and pupils, adopted by governors
  2. A staff acceptable use agreement — what staff can and can't do with AI tools, especially around pupil data
  3. A governor briefing — so the governing body understands what it's approving and can evidence oversight
  4. A parent letter — explaining the school's approach in plain English, before parents ask

Together, these show a school that has thought it through — rather than one scrambling after a data breach or a parent complaint.

Your KCSIE 2026 AI Action Checklist

Work through these this half-term:

Want the whole set done for you — free?

We're drafting the complete KCSIE-2026-ready document set — AI policy, staff agreement, governor briefing and parent letter — free for a small number of schools. Thirty minutes of your time. No catch: you get compliant, we get a case study.

The Most Common KCSIE 2026 Gaps in Primary Schools

Having worked with primary schools on exactly this, the same gaps come up again and again:

If two or more of those sound familiar, your school is normal — and fixable in a fortnight.

How AskColin Can Help — Free for a Small Number of Schools

AskColin provides AI training and support for UK schools. We've built the complete KCSIE-2026-ready document set with schools already — and right now we're offering to do it free for a small number of schools.

Here's how it works:

  1. Book a 30-minute call. We talk through how your school currently uses AI, your context, and your priorities.
  2. We draft the full set. AI policy, staff acceptable use agreement, governor briefing and parent letter — tailored to your school and aligned to DfE guidance, KCSIE 2026 and UK GDPR.
  3. You take ownership. Everything arrives ready for your headed paper. Edit as you see fit, take it to your governing body, and publish once signed off.

No catch: you get KCSIE 2026 off your list; we get a case study. Thirty minutes of your time gets you a detailed draft of every document your school needs.

Email neil@askcolin.uk or call or text 07971 775986 to get started — or use the contact form and mention KCSIE.

FAQs: KCSIE 2026 and AI in Primary Schools

When did KCSIE 2026 come into force?

1 September 2026, replacing the 2025 edition. Schools are expected to be working to it now.

Does KCSIE 2026 ban AI in primary schools?

No. It expects schools that use AI to understand the safety considerations and legal responsibilities, and to manage risks like deepfakes and AI chatbots within their safeguarding arrangements.

Do primary schools need an AI policy in 2026?

There's no statutory requirement for a standalone document, but AI must be reflected across your safeguarding approach — and a clear, governor-adopted AI policy is the practical way to evidence that.

Can teachers use ChatGPT for school reports?

Only within a clear policy and staff agreement — and never with identifiable pupil data in public AI tools, which risks a UK GDPR breach.

Should our AI policy be on the school website?

It's not required, but publishing it supports the transparency KCSIE expects with parents, and schools already publish comparable policies (online safety, acceptable use) as standard.

How quickly can we get compliant?

With the audit done and the document set drafted, most primary schools can take a full package to their next governors' meeting — realistically two to four weeks.

← Back to all articles Book your free 30-minute call →