Context Window Leakage

Everything an AI system needs for a task sits in its context window: instructions, retrieved data, tool results, and conversation history. Anything in that window can potentially be repeated, summarized, or sent elsewhere.

Common leakage sources include:

• System prompts: Internal instructions and business logic revealed to users.
• Retrieved data: RAG results the user should not be allowed to see.
• Secrets: API keys or tokens placed in prompts or tool outputs.
• Cross-session data: Information from other users or tasks carried into context.
• Tool results: Sensitive API responses echoed back in replies.

Leakage often happens without any attacker: an assistant includes a sensitive detail because it seemed relevant. Prompt injection then makes deliberate extraction far easier.

Reducing leakage starts with keeping secrets out of context, enforcing access controls before retrieval, and inspecting outputs for sensitive content.

How PointGuard AI Helps

PointGuard AI AI Data Protection detects sensitive data in prompts, context, and responses, and AI Runtime Guardrails block attempts to extract system prompts or hidden data. AI Security Posture Management flags applications that place secrets in model context.

Learn More

• OWASP Top 10 for LLM Applications
• OWASP Top 10 for Agentic Applications
• NIST AI Risk Management Framework