What are safety alignments and rule-based prompting used for?

• A. They evaluate model performance after deployment.
• B. They enable the model to memorize restricted content.
• C. Safety alignment ensures outputs adhere to policies; rule-based prompting encodes constraints and guardrails.
• D. They primarily increase token usage efficiency.

Answer: (C)

Safety alignment and rule-based prompting are about shaping model behavior to stay within approved boundaries and constraints. Safety alignment involves designing and tuning the system so outputs follow policy, avoid harmful or inappropriate content, respect privacy, and refuse unsafe requests when needed. Rule-based prompting encodes explicit constraints and guardrails directly into the prompt or system message, such as prohibiting certain topics, enforcing a safe tone, or limiting how information is presented. Together, they steer the model toward compliant, safe responses rather than influencing performance after deployment or focusing on memorization or efficiency.