Defending Large Language Models against Jailbreak Exploits with Responsible AI Considerations

Large Language Models LLMs remain susceptible to jailbreak exploits that bypass safety filters and induce harmful or unethical behavior. This work presents a systematic taxonomy of existing jailbreak defenses across prompt-level, model-level, and training-time interventions, followed by three...