Red Teaming Guides

Practical tutorials for finding vulnerabilities in LLM applications before deployment. Each guide walks through a real-world scenario with working configuration examples you can adapt to your own use case.

Getting Started

• How to Red Team LLM Applications — End-to-end guide to adversarial testing
• Testing Guardrails — Evaluate content safety guardrails

By Application Type

• RAG Red Teaming — Test retrieval-augmented generation for prompt injection, context manipulation, and data poisoning
• Agent Red Teaming — Test LLM agents for privilege escalation, context poisoning, and memory manipulation
• MCP Security Testing — Secure Model Context Protocol servers through red teaming and tool poisoning tests
• Red Teaming a Chatbase Chatbot — Security test a production chatbot

Specialized Testing

• Multi-Modal Red Teaming — Test image and multi-modal AI systems
• Foundation Model Red Teaming — Assess foundation model security risks through red teaming and static scanning
• Evaluating Safety with HarmBench — Benchmark LLM safety using HarmBench
• Testing Google Cloud Model Armor — Assess Google Cloud's AI safety layer

Ongoing Security

• LLM Supply Chain Security — Detect trojans, backdoors, and safety regressions in your LLM supply chain
• Detecting Model Drift — Monitor LLM security posture over time by running red team tests repeatedly