Syllabus AI statement — template
Copy, adjust the bracketed parts, and paste into your syllabus. This is a starting point — your faculty’s academic-integrity policy always takes precedence. Adapted from the kit’s principles; the per-assignment version is the assignment AI statement.
Generative AI in [COURSE CODE]: how we’ll work with it
The short version: this course treats generative AI the way we treat calculators, translators and editors — each task says exactly what is allowed, and you say honestly what you used.
The five levels (from the AI Assessment Scale, v2.1). Every assignment in this course names its level for each stage:
| Level | Name | In this course it means |
|---|---|---|
| 1 | No AI | No generative AI use at all for that part |
| 2 | AI Planning | AI may help you plan or brainstorm before you start the real work |
| 3 | AI Collaboration | AI may give feedback and suggestions during the work; you decide, verify, and log what you took or rejected |
| 4 | Full AI | AI may generate freely; your logged judgment of its output is part of what’s graded |
| 5 | AI Exploration | Working with AI is the task itself |
How to disclose. Each assignment ends with a short AI-use statement: which level each part used, which tool, what you kept, rejected or verified, and one thing you learned. A template is attached to every assignment brief.
What happens at each level. Levels 2–5 are legitimate when disclosed — you lose nothing for honest AI use at the stated level. Undisclosed use where AI is not allowed is handled under [the university’s academic integrity policy — link your policy here]: the concern is the concealment, not the tool.
Why no AI detectors. We don’t use them: published testing found none accurate enough to base an accusation on (none above 80% accuracy; sharply worse on edited and machine-translated text), and they misfire disproportionately on multilingual writers. Your honest disclosure is worth more than a detector score — and asking questions about your work is always how we resolve doubts.
Where AI doesn’t help you. The parts of each task that build your capability are marked No AI for a reason. Bypassing them with AI doesn’t break a rule so much as skip the practice the assessment exists to certify — and the secure checks (in-class, oral, or written defence) will show it.
Access and equity. No assignment requires a paid AI subscription: every AI-permitted level can be completed with free tools or the university’s licensed tool [name it]. If you have no AI access, or an accommodation affects any part of this policy (including oral assessments), contact [instructor email / accessibility office] — alternatives are built in, not negotiated case-by-case.
Questions about AI use on a specific task? Ask before you submit: [instructor email / office hours].
Template: CC BY 4.0, AI-Resilient Assessment Kit. Adapt to your institution’s policies; do not present this as official policy language without that review.