01
Prompt Injection Field Guide
Understand direct and indirect injection, attack paths, and layered defenses.
Read overview →Practical intelligence for safer AI.
Use these starter guides to understand common AI risks and build a more resilient security program.
02
AI Asset Inventory
A practical checklist for finding models, agents, data stores, and third-party AI services.
View checklist →03
Agent Security Basics
Reduce risk when AI can browse, send, purchase, modify, or execute.
Explore controls →04
Governance Starter Map
Connect technical controls with ownership, policy, evidence, and oversight.
Build a map →Prompt injection overview
Prompt injection attempts to influence an AI system with malicious or conflicting instructions. Strong defenses combine trust boundaries, input isolation, constrained tool permissions, output validation, monitoring, and human approval for sensitive actions.
AI inventory checklist
- Models and model providers
- Prompts, system instructions, and retrieval sources
- Agents, plugins, tools, APIs, and service accounts
- Vector databases and sensitive data paths
- Owners, business purpose, users, and deployment environments
Agent safety controls
Give each agent the least privilege needed, validate tool parameters, separate untrusted content from instructions, limit irreversible actions, and require human review for high-impact workflows.
Governance starter map
Assign accountable owners, define acceptable use, document exceptions, establish release gates, test incident response, and retain evidence that controls are operating as intended.