Adversarial AI Security Lab: LLM red-teaming, RAG poisoning attacks, automated evaluations, and defensive mitigations. Includes Garak-based vulnerability scanning, retrieval hijacking demos, and security-focused agent simulations.