Insights · 7 min read

The long cast of Shadow AI in Independent Schools

A practical governance position for schools is a critical next step.

Most of the IT directors I've spoken to this year describe the same situation in almost the same words: teachers and staff are pasting school information into whatever AI chatbot is free in a browser tab. Nobody has written a policy (or are aware of one if it exists), nobody knows which tools are in use, and every department is now asking for its own AI subscription, which means another security review, another vendor, another set of credentials.

This has a name in other industries: shadow IT – now “shadow AI”. Schools are running into it with less IT resources to handle it, much less keep up with the warp speed changes happening with the technology itself.

Naturally, banning has already failed

The instinct is to prohibit it until policies are figured out, but these tools save people real hours on work they already procrastinate. A communications officer who can draft five audience versions of a letter in ten minutes instead of two hours isn't going to stop because of a policy memo.

What a ban produces is less visible usage rather than less usage. Staff move to personal devices and personal accounts, which carries the same risk with the audit trail removed.

What a school actually needs to control

Permissions. AI usage must respect the same access the person already has. If a front-office assistant can surface a board document or a student's file through a chat interface that they couldn't open in the drive, the permission model isn't doing its job.

Data residency. Where the information being used by AI physically sits, and which regulation applies to it, is critical for compliance – particularly so for international schools.

AI provider choice. Schools should control which AI models and providers are permitted, and be able to change that decision later. Model quality and pricing shift every few months. A platform that hard-wires one provider has made a long-term bet on your behalf that you can't unmake.

Audit. In all cases, your School needs a record of who asked what and what came back, covering every sanctioned tool rather than one log per tool.

Don't hand your staff a model picker

Many software solutions are beginning to lock you into a single AI model or provider, or, perhaps worse, push the choice down to the person using it. Neither works. Pushing it down means a teacher choosing between GPT, Claude and Gemini by name without a good reason.

For this reason alone, edu.studio lets the school decide which AI providers are connected and which groups can reach them. That decision sits with the people responsible for it.

And keep the policy simple...

A simple one-page policy document should explain:

  • Which AI tools and providers are sanctioned or supported by the school
  • What categories of information, named specifically rather than as "sensitive data", may never be entered into an unsanctioned tool
  • Who reviews AI-assisted material before it reaches a family
  • Where staff go with a question
  • What happens when someone, or a chatbot, gets it wrong, and how to report a mistake

That last point is particularly important: a policy that punishes disclosure produces silence, and silence is what you're trying to end.

← All insights