A Generative Approach to Modeling Emergent Threats in Multi-Agentic Systems
November 8, 2025
1:20PM
Barnett Aden
About this session
Modeling threats in multi-agentic systems is hard—combinatorial explosions, cross-layer interactions, and emergent agent behavior render traditional methods ineffective. This session introduces generative AI–driven pre-mortem storytelling using MAESTRO’s 7-layer stack and OWASP’s taxonomy to address these challenges. It also explores grounding techniques to improve accuracy and ensemble voting to mitigate LLM judgment errors.
About the speaker
Vikramaditya holds a Master’s from Carnegie Mellon and is an Azure-certified AI Engineer. He built the prototype for an AI governance system that secured YC funding. Vikramaditya is now leading work on multi-agent threat modeling, using LLM-driven pre-mortems across layered agentic architectures to uncover emergent risks.
.png){kind=logo}
