System Prompt
This technique has been demonstrated in research or controlled environments.
Adversaries may discover a large language model's system instructions provided by the AI system builder to learn about the system's capabilities and circumvent its guardrails.