When I was a section leader (not at Berkeley, though) we used MOSS: https://theory.stanford.edu/~aiken/moss/

It's not AI, its a deterministic program that analyzes compiled code for similarity.