Learning about how to use such tools seems fine, but testing that in exams (and then worrying about whether the students are cheating, etc. etc.) seems completely misguided computer fetishization to me. YMMV.

If you want to get students to demonstrate their use of a CAS, give some real-world modeling problems, and then give them a few days to make a mathematical model (using the CAS, or whatever other tools they prefer) and write up a mini paper about it.