Two risks deserve open discussion with your class. First, AI reflects bias in its training data, so its output can disadvantage or misrepresent some groups, and students should learn to read it critically rather than trust it. Second, over-reliance is its own harm: a student who lets a tool think for them does not build the skill the course exists to teach.

Name both directly. Frame AI as a fallible assistant whose work must be checked, and design tasks, as the earlier lessons describe, where the student’s own reasoning is what is assessed. Talking about these risks openly does more good than pretending the tool is either magic or forbidden.