The AI Policy Your Senate Will Actually Approve

By Midya U · Midya U Advisory

In brief: Campus AI policies stall when they try to rule on everything at once. What passes Senate is one short document with five jobs: name the allowed uses, draw the red lines, keep a human on decisions about people, say when AI use is disclosed, and build a fast approval lane. Draft it with all five offices in one room, pilot for a term, bring Senate the evidence.

By 2026, most Canadian campuses have an AI policy "in progress." Far fewer have one in force. The gap between those two states is where the risk lives.

Everyone cares and different departments own a piece — that's part of the problem. IT owns security. Privacy owns student data. Procurement owns the vendors. Legal owns the liability. Senate owns academic integrity. Each one can stall a draft, and most do. I've watched a policy circulate for over a year while the version number climbed and nothing shipped.

Meanwhile, the work doesn't wait. Your staff are already using AI — a dozen tools across the org, most of them never formally approved, because the report is still due Friday. While the draft circulates, the institution carries the risk without any guardrails.

Why do campus AI policies stall in committee?

The failed policies I've read share a shape. They're written as a catalogue of risks, usually by one office, usually the one that fears AI most. They run twelve to twenty pages. They try to rule on everything at once — teaching, research data, admissions chatbots, marketing copy. And every office reads the same paragraph differently, which means every review cycle produces new objections instead of fewer.

A document like that asks Senate to approve everything about AI in one motion — and no Senate will.

What should a university AI policy include?

The policy that gets adopted is the one a professor, a marketer, and a lawyer can all read in ten minutes and apply the same way. It does five jobs, in plain language:

1. Name what's allowed. Real examples of sanctioned use — drafting, summarizing, translation, first-pass analysis — with the approved tools listed by name.
2. Draw the red lines. Student and employee personal information never goes into public AI tools. Whatever else your provincial privacy legislation requires sits here too, stated once, clearly.
3. Keep a human on decisions that affect people. Admissions, grades, financial aid, hiring. AI can prepare the file; a person signs it.
4. Say when AI use gets disclosed. To students, to applicants, to the public. Leave this ambiguous and it will cost you trust with all three.
5. Build the fast lane. A route for approving a new tool in weeks, with a named owner. If the sanctioned path takes two terms, people will route around it, and you're back where you started.

How do you get an AI policy through Senate?

Get the five offices in one room and draft together — one working day beats a year of circulation. Pilot the policy in a single unit for a term. Then bring Senate the pilot's evidence instead of hypotheticals: what people used, what got flagged, what the guardrails caught. Committees argue with predictions; they have a much harder time arguing with a term of data. Put a review date on the policy before anyone asks for one — a policy you'll revisit in a year doesn't have to settle everything today.