A is correct: Negative constraints are employed to direct the model away from specific unwanted behaviors, themes, or stylistic elements.

B is incorrect: While a prompt can instruct a model to stop, negative prompting specifically defines what *not* to do or include, rather than just stopping generation.

C is incorrect: Negative prompting is not about instructing the model to be 'mean'; it's about defining undesirable outputs.

D is incorrect: Negative prompting relates to content generation directives, not to the model's power consumption.