How schools can build practical AI rules instead of blanket bans

Why blanket rules are difficult

Generative AI can be used for brainstorming, tutoring, translation, editing, research, coding or producing an entire assignment. Treating all of those activities as identical makes a policy hard to apply.

A better approach starts with the learning objective. If an assignment is designed to assess a student's independent writing, generating the entire response is different from using a tool to identify spelling errors after the student has written a draft.

UNESCO's human-centred approach

UNESCO's guidance on generative AI in education recommends a human-centred approach and highlights issues including privacy, age-appropriate use, teacher preparation and the need for coherent policy frameworks.

That is useful because schools are not only managing cheating. They are also deciding how students should learn to work with a technology that will exist beyond the classroom.

Build rules around categories of use

A practical policy can define categories such as: prohibited use, allowed assistance, disclosure-required use and teacher-approved use. Each category should include examples.

For example, a school might permit AI brainstorming but require the student to document how the tool was used. Another assignment might prohibit generative writing because the purpose is to measure the student's own composition skills.

Privacy should be part of the policy

Students may enter personal information, unpublished work or sensitive school data into external tools without understanding what happens next. Schools should explain what information must never be submitted and which approved tools, if any, may be used.

Teacher training matters too. A rule is only useful if the people applying it understand the tools and can explain the expectations consistently.

Assess the work, not just the tool

AI detection alone is a weak foundation for an education policy. A stronger assessment model can include drafts, oral explanations, classroom writing, source checks and process evidence where appropriate.

The goal is not to pretend AI does not exist. It is to make sure students still develop the knowledge and skills the course is intended to teach.

Start with the learning objective

The same AI action can be appropriate in one assignment and inappropriate in another. If a class is testing independent writing, generating the final essay may defeat the purpose. If the class is teaching editing, using AI to identify grammatical patterns might be useful.

This is why blanket rules are difficult to enforce. A policy becomes clearer when teachers first define what students are expected to learn and then decide where AI assistance supports or undermines that goal.

Create simple use categories

Schools can define categories such as prohibited use, permitted assistance, disclosure-required assistance and teacher-approved use. Each category should include concrete examples rather than abstract language.

For example, brainstorming may be allowed while generating a final answer is prohibited. A student might be allowed to use a tool for proofreading but required to submit the original draft. Exact rules should reflect the course and age of the learners.

Privacy should be explicit

Students may not understand that entering personal information, unpublished work or sensitive school data into an external AI service can create privacy implications. Policies should identify what information should never be submitted and which tools, if any, are approved.

Teachers also need guidance. A policy that students cannot understand and teachers cannot apply consistently will create confusion rather than trust.

Assessment should show learning

AI detection tools should not become the sole foundation of academic-integrity decisions. A stronger approach can include drafts, classroom work, oral explanations, citations, project history and other evidence appropriate to the subject.

The objective is to evaluate whether the student developed the intended knowledge and skills. That can sometimes mean redesigning assessments so the process matters as much as the final artifact.

• Define the learning objective.
• State allowed and prohibited AI uses.
• Require disclosure where AI assistance matters.
• Protect personal and sensitive information.
• Give teachers examples and escalation guidance.
• Review the policy as tools and classroom practices change.

Questions readers ask

Should schools ban generative AI?

A blanket ban can be difficult to enforce and may not teach students how to use AI responsibly. Some assignments may need strict limits, while others can use AI as part of learning.

How should students disclose AI use?

The method should be simple and proportional. A teacher can specify whether students should name the tool, describe how it was used or provide prompts and outputs for certain assignments.

How can schools protect students?

Use approved tools where possible, teach students not to submit sensitive information and make privacy expectations part of digital literacy.

The strongest policy is not the longest policy. It is the one a student can understand, a teacher can apply and a school can update as technology changes. AI literacy and academic integrity can coexist when expectations are explicit.

Build rules around learning goals

A school policy should begin by defining what a particular assignment is meant to assess. If the goal is to practice drafting, unrestricted generated writing may undermine the exercise. If the goal is to compare sources or critique a model's output, AI use may be a useful part of the assignment. The same tool can be acceptable in one context and inappropriate in another, so a single blanket rule often leaves students and teachers guessing.

Teachers can state allowed uses in plain language: no AI assistance, brainstorming only, language feedback permitted, or AI use required with a short reflection. When disclosure is expected, explain what counts as meaningful disclosure and give students a simple format. A consistent rule across a course is easier to apply than a policy that changes informally from task to task.

Protect privacy and keep assessment fair

Students should not enter sensitive personal information, confidential school records or identifiable work from classmates into tools that have not been approved for that purpose. Schools should consider age requirements, data handling, account creation and whether a tool uses submitted material to improve its services. Teachers should avoid uploading student work into public AI detectors as if those systems can conclusively prove authorship; false positives can damage trust and require a fair review process.

Assessment can be made more robust by including drafts, oral explanation, in-class work, reflections on decisions and opportunities to discuss sources. These methods provide evidence of learning without turning every assignment into a contest between students and detection software. Students should also learn to verify factual claims, identify fabricated citations and recognize the limits of generated answers.

A good policy is teachable and revisable. Invite feedback from students, teachers, families and safeguarding staff, then review incidents and classroom experience at set intervals. The objective is not to endorse every AI tool or ban every use; it is to preserve learning, fairness, privacy and clear expectations as tools change.