AI Red-Teaming: Defeating LLM Security Controls

An engineer's guide to prompt injections, indirect injections, state jailbreaking, and how to verify robustness against adversarial model manipulation.