Can I Trust My AI? Securing Agentic Intelligence in a Risk-Driven World

AI is evolving from predictive models to autonomous, agentic systems capable of decision-making with minimal human oversight. This shift introduces new security challenges and raises critical questions: Can we trust our AI—and should we? The talk examines the risks of agentic AI, the role of emerging standards like Model Context Protocol (MCP), and vulnerabilities across machine learning attack surfaces. We contrast trust with formal security guarantees, explore compliance frameworks such as the EU AI Act, and propose an approach to design, development, and deployment for reliability, resilience, and responsibility.