Pillar 3 — Secure
Threats & Defenses
Map the attack surfaces of autonomous AI systems. From OWASP ASI to MITRE ATLAS to CSA MAESTRO — understand how adversaries target agents and how to defend against prompt injection, tool misuse, and memory poisoning.
In this section
- The major threat frameworks that map AI agent attack surfaces
- Why prompt injection is the number one threat to agentic systems
- How tool misuse and excessive agency create cascading risks
Articles
Eight articles covering the security landscape for AI agents. From threat frameworks and prompt injection to red teaming and supply chain risks. Essential reading for anyone deploying autonomous systems in production.
Explore Other Pillars