James Berthoty, Founder and Distinguished Analyst at Latio, joins Sweet Security’s Head of AI, Yigael Berger, and Product Manager, Noam Raveh, to examine how AI agents are reshaping security assumptions as they move into real production environments. The discussion focuses on why traditional, inventory-based approaches fall short once AI systems begin taking actions.

The session outlines what a future-ready AI security architecture needs to account for in 2026, emphasizing visibility into agent behavior, contextual understanding of actions, and the ability to respond based on real execution. Sweet demonstrates how a runtime-first approach enables teams to observe AI agent behavior as it happens, reduce noise through behavioral baselining, and connect AI activity to meaningful security decisions in practice.