ai-teaching-guardrailslisted

# Working with AI on teaching and academic work Apply these whenever the person will put their name on the output. The `course-audit` and `course-rebuild` skills assume these rules; this skill states them, and stands on its own for anything else. ## Two rules underneath all the others ### A. Whatever you tell students about using AI applies to the teacher as well If a course tells students to do the work by hand first, then audit the AI rather than outproduce it, then disclose what they used, the teacher cannot skip all three. Not because it is unfair — because it does not work. A lecture written by an AI is not owned by the person delivering it, and that shows the first time a student asks something off-script. ### B. Assume every student has an AI assistant, in everything they do Not "might have". Has. Design from that assumption, and two things follow. **A ban is never a control.** If part of a course is meant to be done without AI, its integrity has to come from the design — done in the room, on data the student has not seen before, with a short closed-book piece at the end — never from an instruction and an honour system. Write the rule anyway, because it tells students what the task is for. Just do not rely on it. **Learning with AI is part of what you are teaching, not a threat to it.** Design tasks that make students better at using it: get an answer from it and find what is wrong with that answer; ask it for the strongest case against your own position; hand