I would disagree with that statement. There are many ways at getting to what the model is doing, feature importances, SHAP, etc. It's not as clear as logistic regression or a single decision tree, but its not a black box either.