How to Use AI Safely in Schools: A 2026 Guide for Educators

The debate about whether schools should use AI is largely over. The question now is how. Done well, AI can support learning and save teachers time. Done badly, it can expose student data, show children inappropriate content, and replace thinking instead of building it.

To use AI safely in schools, choose a platform designed for your students' ages, get a written data processing agreement before any student uses it, enforce age-appropriate boundaries in the platform itself, make sure teachers can review student conversations, teach AI literacy, write a short policy, and pilot in one classroom before scaling. This guide walks through each step, mapped to the five risks every school needs to manage, with links to current UK and US guidance.

The Five Risks Every School Needs to Manage

Every safety step below maps back to one of these:

1. Privacy and data: student conversations, personal information, academic records.
2. Content exposure: age-inappropriate answers, and errors presented as fact.
3. Misuse: AI-written assignments and plagiarism.
4. Dependency: students offloading thinking instead of developing it.
5. Equity: uneven access between students with AI at home and those without.

A school's AI safety plan is simply concrete answers to those five.

What Current Guidance Says

• England: The Department for Education's Generative AI: product safety standards, updated on 19 January 2026, set out what AI products used in schools and colleges should meet. The 2026 update added standards on cognitive development, emotional and social development, mental health and manipulation. They are written for suppliers, but schools can use them to assess products.
• UK data protection: The ICO's Children's Code applies to online services likely to be used by children.
• United States: In July 2025, the Department of Education issued guidance on using federal grant funds for AI, stressing compliance with federal privacy law and involving parents in adoption decisions. Student records are covered by FERPA, and online services for children under 13 by the FTC's COPPA Rule.

1. Pick a Platform Built for Your Students' Ages

This is the highest-leverage decision you will make. Consumer chatbot accounts are not designed for young students: ChatGPT's terms require users to be at least 13, and Anthropic's consumer Claude service is for adults. Education editions are a step up. Google made the Gemini app available to Workspace for Education users of all ages in June 2025, with stricter content policies for under-18s and admin controls, and Microsoft says students aged 13 and older can use Copilot Chat. They remain general-purpose assistants, though, not tools designed around a seven-year-old.

For younger students, look for a platform built for children, with:

• Explicit COPPA compliance and a signed data agreement
• Answers adapted to the student's age
• Teacher visibility into student conversations
• No advertising and no sale of student data

See our top 10 AI tools for schools comparison for a side-by-side.

2. Get a Data Processing Agreement Before Students Use It

Before any AI tool sees a student's name, question or work, the school should have in writing:

• What data is collected, and from whom
• Where it is stored, and how it is encrypted
• Whether it is used to train models (for student data, the answer should be no)
• How long it is kept, and how to request deletion
• Sub-processors: who else handles the data
• Breach notification: how quickly you will be told

In the US, this maps to FERPA and, for under-13s, COPPA. In the UK, to UK GDPR and the Children's Code. In the EU, to GDPR and national education law. Vendors that are ready for schools have these answers prepared.

3. Set Age-Appropriate Boundaries, Enforced by the Platform

A seven-year-old and a fourteen-year-old should not have the same AI experience.

• Ages 5-7: voice or read-aloud; tightly scoped topics; always with a teacher present. The goal is wonder, not autonomy.
• Ages 8-10: guided exploration within teacher-set boundaries; conversations visible to the teacher.
• Ages 11-13: broader scope inside a school-managed platform; explicit teaching about AI's limits.
• Ages 14-18: more autonomy, with explicit teaching on academic honesty, citation and how AI fails.

The crucial point: these boundaries should be enforced by the platform, not left to "rules we told the class". Children will not consistently self-regulate around an AI that can answer anything.

For primary-specific guidance, see AI for elementary schools.

4. Make Every Interaction Reviewable

A simple test for any classroom AI: can a teacher see, after the fact, what a student asked and what the AI replied?

If not, the tool is hard to justify. Not because teachers will read every line, but because:

• A safeguarding concern may require looking back at a conversation.
• A parent may ask what their child was doing.
• A disputed answer in assessed work needs an audit trail.

In England, the DfE's product safety standards go further: learner-facing products are expected to alert local supervisors to harmful content and to disclosures that suggest a safeguarding issue, and to send high-risk alerts to the designated safeguarding lead. Ask every vendor exactly how that works in their product.

5. Teach AI Literacy, Not Just AI Use

Students who understand how AI fails use it better than students who only know how to prompt it. Build these ideas into existing lessons:

• AI makes things up. It can state false facts confidently. Always verify.
• Disclose AI use. If you used AI, say so, and set that norm early.
• Prompts shape answers. The same question phrased two ways gets different answers.
• AI predicts, it does not understand. Useful framing for older students.
• Bias is real. AI inherits the biases of its training data.

A few minutes a week, woven into existing lessons, changes how students relate to the tool.

6. Write a One-Page School AI Policy

Long policies rarely get finished. A one-page policy gets adopted. It should answer:

1. Which AI tools are approved for which year groups?
2. What is acceptable use for students?
3. What is acceptable use for staff (planning, feedback, admin)?
4. How is AI use disclosed in student work?
5. What happens when the policy is broken?

Our AI policy for schools template is a ready-made starting point.

7. Pilot Before You Scale

A common failure is not a safety incident but a top-down rollout that teachers quietly ignore. The fix is to pilot first: one classroom, one teacher, six weeks, with success criteria agreed at the start (engagement, time saved, learning, teacher confidence). Write up what worked, then decide whether to scale. Our guide on how schools can pilot AI is a step-by-step companion.

8. Plan for Equity From Day One

If some students have paid AI tools at home and others do not, AI can widen gaps rather than close them.

• Make in-school AI available to every student, not just selected classes.
• Provide structured AI time in the school day so home access matters less.
• For students with SEND or English as an additional language, well-chosen AI can be especially helpful. See AI for special education.

9. Have a Plan for AI-Assisted Cheating

You will not catch every instance, and trying to can create an adversarial classroom.

• Redesign assessments. Drafts, conferences and oral presentations make the process visible.
• Be explicit about which tasks allow AI, and what disclosure looks like.
• Use AI detectors cautiously. A Stanford study found that popular AI detectors often misclassified writing by non-native English speakers as AI-generated. See does AI detection in schools work? for more.

10. Communicate With Parents Early

Parents will hear about your AI rollout one way or another. It is better that it comes from you. A short letter home should cover:

• Which AI is used, in which years, and for what
• The safety and privacy position in one paragraph
• How parents can see what their child is doing
• A named contact for questions

The Honest Summary

Safe AI in schools is mostly unglamorous: signed data agreements, age-appropriate platforms, teacher visibility, a one-page policy, and a pilot before anyone scales. There is no single safety switch. There is the steady work of choosing well and writing down what is expected.

Frequently Asked Questions

Is AI safe to use in schools?

It can be, when the platform suits the students' ages, the school has a signed data agreement, teachers can review conversations, and boundaries are enforced by the platform. Consumer chatbot accounts are not designed for young students.

What is the biggest mistake schools make with AI?

Rolling out a tool across the school or district before any teacher has piloted it. A short classroom pilot builds trust and shows what actually works.

Does AI in schools break student privacy laws?

Not inherently. Problems arise when schools use consumer tools without proper agreements, when student data is used to train models, or when under-13s use services that are not COPPA compliant.

Should schools ban ChatGPT?

Bans rarely work, because students can use it on personal devices. A better approach is to provide an approved, age-appropriate tool, teach AI literacy, and redesign assessment. See ChatGPT in schools: should students use it?.

How do we get started?

Pilot in one classroom with a platform suited to your students' ages, sign the data agreement first, run for six weeks, then decide whether to scale.

---

Ready to roll out AI safely in your school? Askie for Schools is a separate classroom deployment of Askie with teacher dashboards and student access codes, built on the same multi-layer safety filtering and age-adaptive answers for ages 4-15. The free plan covers 20 students across your classes, free forever, with 200 text questions each week and no credit card required. Start a pilot