People make mistakes, too.
That's the nature of intelligence, "artificial" or not.
What's surprising to some people, though it shouldn't be if you spent more than a couple seconds thinking about it, is the mistakes AI makes are different than the mistakes humans make.
As far as "messing up" is concerned, I'm more concerned about my car attempting to drive me into a train than I am with Gemini returning an obviously absurd result that I can discard.