LLM Safety, Security, Privacy, and Red Teaming

Learning objectives

• • Threat model RAG, tool-using agents, and external integrations
• • Design red-team tests and human escalation paths
• • Balance logs, privacy, retention, and audit requirements

Interview prompts

• • How do you defend a tool-using agent from prompt injection?
• • What should trigger human review in a high-risk LLM workflow?

Prerequisites