Prompt injection defence
Limiting what untrusted content can make an agent do.
Articles
-
A map of AI agent governance techniques
Thirteen families of controls for AI agents, where each one sits in the request path, and which risk it actually reduces.