Click to add a weight penalty to the loss. Large weights now cost extra, so the model must balance fitting the data against keeping its weights small.
Regularization adds a second term to the loss: a charge for having large weights. The model now has to balance fitting the data against keeping itself small.
