Not quite. You're appleaing to the Curry-Howard correspondance, which is neat but is not used by many theorem provers. Furthermore,
> fun fix (f : 'a -> 'a) : 'a = fix f;
val fix = fn: ('a -> 'a) -> 'a
The trick is that all functions must be total, which is the crux of the matter (and why the Y-combinator exhibits the paradox, because it allows nontermination). This is why we have stuff like the polymorphic lambda calculus, in which termination is guaranteed and the Y-combinator is untypeable.
In such a language, types classify terms and types correspond to theorems. The type/theorem for a fixed point is
which as a theorem reads: "for any proposition A, if we can prove 'if A, then A' then we can prove A". Hopefully it's now clear why having 'fix' in your prover language won't work!