Speaker: Najat Diamante

In an era where AI systems autonomously execute workflows without human intervention at each step, traditional oversight mechanisms become a mere illusion. This presentation illuminates the specific risks inherent in agentic behavior-from prompt injection attacks and unexpected system behavior to unauthorized API access. Simultaneously, it demonstrates how modern AI governance must be architected to preserve accountability and traceability even under maximum automation. Participants will learn practical governance approaches, monitoring strategies, and regulatory requirements necessary to operate agentic systems safely and compliantly-while avoiding the trap of false control illusions