Jailbreaking

LLM jailbreaking lets attackers bypass AI safety controls using nothing but a crafted prompt — no CVE required. Learn how to detect jailbreak attempts in your logs, test your own models with Garak, and layer defenses that actually hold.