Prompt framing is the way a question or instruction is worded, and it can strongly influence how an AI model responds. Small changes in tone, assumptions, or structure can push a model toward agreement, caution, detail, or refusal.
In cybersecurity, prompt framing matters because attackers and defenders both use it. An attacker may phrase a request to bypass safeguards, elicit sensitive guidance, or make a model accept a false premise. A defender may frame prompts to demand evidence, ask for alternative views, or force the model to challenge weak assumptions. Good prompt framing does not make an AI system trustworthy by itself, but it can reduce sycophantic answers and improve verification in high-stakes workflows.