In chess a match where both players make optimal moves is generally believed to result in a draw.

With future super intelligent language models, we are going to find out if many other aspects of life are a win or loss with optimal play by both sides.

Software security is one, I predict this is probably a win for the defender who can harden the code against all possible attacks. Cryptography, on the other hand, might turn out to be always a win by the attacker if the mathematics of asymmetric key crypto falls apart.