Security testing for LLM-powered applications including prompt injection, jailbreak detection, data leakage prevention, and AI safety testing.