HOW TO USE THIS GUIDE
ABOUT THIS COMIC BOOK
Welcome to the AI Security guide focused on Agentic AI systems! This comic breaks down complex security threats into easy-to-understand visuals and explanations based on the OWASP Top 10 for Agentic Applications 2026 — the first authoritative list of security risks specific to AI agents that plan, decide, and act autonomously.
Unlike traditional LLM risks, agentic threats emerge when AI systems operate across multiple steps, use tools, store memory, and delegate to other agents — often without a human in the loop.
PAGE LAYOUT
Each vulnerability is presented across a two-page spread:
LEFT PAGE — Four elements:
A full comic-style scene depicting the threat in action
A real documented incident where this exact threat caused harm — names, dates, and what happened
Plain-English definitions of the key technical terms used on each page
Which OWASP LLM Top 10 items this agentic threat builds upon — bridging the two frameworks
RIGHT PAGE — Four sections:
A simple explanation with a relatable analogy
Step-by-step breakdown of the attack chain
Real-world impact — with CIA markers showing what's at risk
Actionable mitigations you can start implementing today
CIA IMPACT MARKERS
In the Why It Matters section, look for these icons showing which security pillars each threat affects:
Confidentiality
Risk of sensitive data being exposed or stolen
Integrity
Risk of data or decisions being corrupted or falsified
Availability
Risk of systems becoming inaccessible or disrupted
NAVIGATION
- ← Previous page
- → Next page
- ☰ MENU Jump back to the Table of Contents at any time to navigate directly to any topic