[AaronFriel]

Because the model has been trained to do what you tell it to do? That's what instruction pretraining/fine-tuning is.