It isn’t caution to overestimate.
Overestimating the probability of X is just underestimating the probability of not(X).
Overestimating the probability of X is just underestimating the probability of not(X).