When does regularization become necessary in Machine Learning?

A) When the model is underfitting the training data.
B) When the model is overfitting the training data.
C) Only when dealing with large datasets.
D) Regularization is never necessary in Machine Learning.