IN FOCUS
A threat model for agentic systems
Start with input attacks, identity and permission risks, tool and action risks, data risks, and the connections that make MCP useful.
SECURITY
AI Agent Security Threats
Eyrane organizes agentic security problems around what can influence an agent, what it can access, what it can do, and where data can go.