Saltar al contenido principal

Threat Model an AI Agent

Create a practical threat model for an AI agent that names boundaries, threats, controls, and tests.

building-ai-apps5 min read

An AI customer-success agent reads untrusted customer emails and can update CRM opportunity stages. Diagram -> classify -> control -> test The common trap is saying the model will not do dangerous things because the system prompt forbids them. In an agent, the application must constrain tool authority even when the model is confused. Draw the boundary Map user, customer email, retrieval store, model context, CRM tool, secrets, audit log, and CRM database. This shows where low-trust text can influence high-authority actions. Classify threats Apply STRIDE: spoofing user identity, tampering with tool arguments, repudiation without logs, information disclosure from CRM,…

Read the full lesson

Sign up free: one personalized lesson every day, matched to your role and goals.

← Back to library

Contáctanos