Agentic AI security (OWASP Agentic Top 10)
A · 10-second intuition Agents turn model mistakes and manipulations into actions, so agent security is about bounding what any single decision can do.
Predict before you read
An injected document makes an agent pursue the attacker's objective instead of the user's. Which OWASP agentic risk is this?