---
title: Security, Compliance & Limitations
description: Understanding and mitigating the real security risks of AI-assisted development
head:
  - - meta
    - name: keywords
      content: AI security, prompt injection, agentic architecture, security boundaries, threat modeling
---

# Security, Compliance & Limitations

> Security isn't about abstract "AI has risks"—it's about concrete design of execution environments, permissions, and isolation.

This category addresses the real security challenges of AI-assisted development. Not fear-mongering, but practical threat models and defense strategies that help you build safely.

## Articles in This Category

| Article | Author | Key Question |
|---|---|---|
| [AI Agents: Threats and Mitigations](./ai-agents-threats-and-mitigations.md) | Jay Chen & Royce Lu (Palo Alto Networks) | What are the specific attack vectors against AI agents? |
| [Security Boundaries in Agentic Architectures](./security-boundaries-in-agentic-architectures.md) | Addy Osmani | How do you design isolation and permission models for agents? |

## Reading Recommendations

- **Both articles are essential reading** — Start with "Threats and Mitigations" for the threat landscape, then read "Security Boundaries" for defensive architecture patterns