Skip to content

Prompt and Content Attacks

This module classifies prompt injection, jailbreaks, and system prompt leakage, then covers model output handling and defenses against secret exfiltration. For architectural defenses such as Dual LLM, least privilege, and execution isolation, see Agent Security. This module focuses on the taxonomy and additional output-side concerns rather than repeating those designs.

Chapters

  1. Chapter 2: Prompt Injection, Jailbreaks, and System Prompt Leakage
  2. Chapter 3: Output Handling and Secret Exfiltration

Return to AI Security and Governance.