Very interesting game. In the first iterations the AI was fairly predictable but it would be foolish to think you can employ the same strategy every time. Did you use a reinforcement learning algorithm for the AI?
Also there is a delicious irony for the story of the game: the AI claims to be all powerful but can only gain an advantage through powerups and debuffing the opponent.