System prompts form the foundation of generative AI applications, containing proprietary information such as role definitions, behavioral guidelines, and tool descriptions. System prompt leakage is a frequently reported security finding, appearing in the 2025 OWASP LLM Top 10 as LLM07. Leaks occur when a generative AI application discloses its instructions or operational context, often through prompt injection where crafted inputs manipulate the model into revealing portions of the system prompt. Multi-turn extraction techniques can be more effective at gradually bypassing safeguards. In agentic applications using tool calling and multi-step orchestration, any prompt leak can expose tool definitions, schemas, and orchestration logic. Exposure of user-specific information is not a concern because users already have authorized access to their own data. Publicly documented events show researchers have extracted partial or full system prompts from numerous widely deployed generative AI applications, with collections cataloged across multiple public GitHub repositories. System prompt leakage does not currently have a remediation that fully eliminates the issue, as this is a fundamental limitation of current generative AI systems.