This sort of strategy game AI in general works on a Dark Forest-ish theory, recursively evaluating what it would do if it were in the opponent's shoes to predic

This sort of strategy game AI in general works on a Dark Forest-ish theory, recursively evaluating what it would do if it were in the opponent's shoes to predict the reaction it will get to each possible move. Since Go is zero-sum the only reasonable strategy is to be maximally malicious and assume the other guy is too.