How Canadian Universities Are Actually Writing Their AI Policies After ChatGPT

One university, several sets of AI rules

Your school does not run one generative AI policy. It runs as many as it has instructors willing to write one. A September 2025 analysis in Policy Options found that only about half of Canadian universities have any formal policy on generative AI, and most of those push the real decision down to the person teaching the course.

This matters for you directly. The same tool that gets a warning attached to it in one class is a required part of the workflow in another, at the same school, in the same term.

Why schools moved away from blanket bans

When ChatGPT reached students in late 2022 and early 2023, the first instinct at many institutions was a blanket ban written into course outlines or handed down from a dean's office. That approach broke down fast. Instructors in writing-heavy humanities courses wanted stricter limits than instructors in computer science or design courses, where AI tools are already part of the professional toolkit graduates use on the job.

Higher Education Strategy Associates (HESA) runs an AI Observatory , a clearinghouse where Canadian post-secondary institutions compare policies and guidelines with each other. Its existence is itself a sign of how scattered institutional policy is: schools need a shared reference point because they are not converging on one national standard.

The per-course model, in practice: University of Toronto

The University of Toronto shows how the delegated model works. Under its Code of Behaviour on Academic Matters , updated by Governing Council on June 24, 2025 and effective July 1, 2025, using ChatGPT or another generative AI tool on a marked assessment is treated as using an unauthorized aid unless your instructor has explicitly allowed it for that assignment. The default is off. Permission comes from the instructor, assignment by assignment, not from a university-wide green light.

If you are not sure whether a tool is allowed on a specific assignment, the university's own guidance is to ask the instructor directly rather than guess.

Detection tools are unwinding too

The patchwork extends to how schools try to catch unauthorized AI use. The University of Waterloo turned off Turnitin's AI-detection flag starting September 2025 , after internal testing found the tool flagged human-written text as fully AI-generated in multiple cases. UBC has held a similar position since 2023, choosing not to enable Turnitin's AI-detection feature at all, citing unresolved accuracy and bias concerns.

If your school still shows AI-detection scores, treat a flag as a conversation starter with your instructor, not proof of an offence. The tools behind those scores have a documented history of false positives.

The problem with leaving it to each instructor

Delegating the decision to instructors solves one problem and creates another. A professor teaching a 30-person seminar writes a detailed AI statement and checks in with students individually. A course with 400 students split across a dozen teaching-assistant sections struggles to enforce one rule consistently across every section, even when the syllabus states one clearly.

Students who took high school courses where AI use was normalized, or who are more confident testing the edges of a vague rule, end up with more room to interpret ambiguous wording in their favour. Students who play it safe lose time redoing work by hand that a permissive class down the hall does with AI assistance built in.

Is a national standard coming?

Policy Options has argued Canada needs a coordinated approach, pointing to the UK's Russell Group, which published shared principles on generative AI use across its member universities instead of leaving each one to write its own from scratch. Canada has no equivalent body speaking for all universities on AI, and HESA's Observatory functions as a reference library rather than a standard-setter.

Expect the patchwork to hold through the next few admissions cycles.

What this means if you take courses at more than one institution

Letters of permission, exchange terms, dual credit and transfer courses put you under two policies at once. Before you use any AI tool on graded work, do this for each course:

• Read the syllabus AI statement, not only the general academic integrity policy on the school website.

• If the syllabus is silent, email the instructor and get the answer in writing before you submit anything.

• Keep prompt history and drafts if a course allows AI with disclosure, in case you need to show your process.

• Do not assume a rule from one course, one term, or one school carries over to the next.

Sources

• Policy Options: Canada needs a national strategy for generative AI in higher education
• HESA: AI Observatory
• University of Toronto: Using ChatGPT or other AI tool on a marked assessment
• University of Waterloo: Discontinuing AI detection tool Turnitin.com
• UBC Teaching with Technology: UBC not enabling Turnitin's AI-detection