What a school AI policy actually needs to say

Most are copied from another school and describe tools rather than tasks. Four questions your staff need answered before they can use one.

Almost every school is now being asked for an AI policy - by a board, an inspector, or a parent. Most of the ones being written are copied from another school's, and share the same weakness: they describe tools rather than tasks. A policy built around a list of named products is out of date within a term, and useless the first time a student uses something that is not on the list.

Banning it outright does not survive contact with reality either. Students have these tools on their phones, teachers are using them to plan lessons whether or not the policy mentions it, and a rule everybody quietly breaks teaches a worse lesson than no rule at all. The useful question is not whether AI is allowed, but what a person may hand over, and for which piece of work.

That is the first thing the policy has to answer: allowed for what. 'AI may be used to help plan an essay but not to write it' is a rule a fourteen-year-old can follow and a teacher can enforce. 'AI use must be appropriate' is not a rule, it is a hope. Go task by task - research, planning, drafting, checking, marking - and say plainly which side of the line each one falls on.

The second is who checks, and how. If the policy says work must be a student's own, someone has to be able to tell. In practice that means assessment which is hard to outsource - work done in class, work that builds on a specific lesson, a spoken follow-up on a submitted essay - far more than it means detection software, which is unreliable enough that accusing a student on its say-so is a serious risk.

The third is where the data goes. Every tool your staff paste student work into is a place that work now lives. Before a tool is approved, someone should be able to answer three things: which company holds the data, whether it is used to train their models, and whether a parent asking what you hold on their child could be given a complete answer. If the answer to the third is no, the tool is not approved, whatever it does.

The fourth is what happens when someone breaks the rule - and for a first offence the answer should almost always be teaching rather than punishment. A student who had an AI write their coursework has usually misunderstood what the coursework was for. That is a conversation about learning, and treating it purely as cheating skips the part that changes their behaviour next time.

A good policy fits on two pages, names no products, and lets a teacher make a decision in ten seconds without finding someone senior. If yours cannot do that, it will be filed and ignored - which is the same as not having one.