About
Sessions
WHEN AI GOES ROGUE: Jailbroken LLMs, Autonomous Agents, and the New Cyber Threat
What you will learn:
What happens when an LLM stops refusing—and starts acting? This session explores how jailbroken and rogue LLMs, combined with autonomous agents and access to real tools, can transform prompt manipulation into real-world cyber risk. We examine recent incidents, the emerging agentic attack surface, and demonstrate how Imperum Agent Studio can simulate these threats while Virtus Sentinel detects rogue LLMs, prompt injection, jailbreak attempts, and malicious agent behavior—turning the attacker’s AI advantage into a powerful tool for continuously testing and hardening cyber defenses.