That kind of depends how you define autonomy. Whichever way, I'm not sure I get how "virtual" is a better descriptor for implying a lack of it than "artificial" is.
Also by "we don’t actually know how it works" do you mean that we can't explain why a particular "decision" was made by an AI, or do you mean that we don't know how AI works in general? If it's the first that's generally true, if it's the second I disagree (we know a lot, but still have a lot to learn).
Can they not identify the solution unless they're willing to risk their life?