Security and Privacy in Agentic Systems
Overview
Security and Privacy in Agentic Systems is an intensive one-day experience crafted for AI engineers, security architects, and privacy professionals who must safeguard autonomous AI deployments. Participants explore foundational principles—such as stringent identity and permission management, robust encryption, and comprehensive logging frameworks—that establish trust in every action an AI agent performs. The course highlights the unique failure modes in single-agent and multi-agent LLM environments, demonstrating how prompt injection, model misbehavior, and coordination breakdowns can cascade into real-world vulnerabilities. Drawing on leading standards and cutting-edge privacy-enhancing technologies like differential privacy and homomorphic encryption, attendees learn to design secure architectures and
sandboxed workflows that protect sensitive data at each processing stage. Through practical labs designed to give participants hands-on experience in securing agentic AI systems, professionals will develop the skills needed to anticipate threats, harden agent interactions, and maintain compliance in rapidly evolving regulatory landscapes.
Duration
1 day
Who Should Take This Course
Audience
Security Professionals, Data Scientists/Engineers, AI/ML/Sec/Dev-Ops & SRE Staff, Devs, Managers.
Prerequisites
Participants must have a basic-to-moderate understanding of LLMs and have a computer capable of logging into a cloud lab system via ssh. Basic Linux command line skills and some coding experience are helpful but not required.
Why You Should Take This Course
Security and Privacy in Agentic Systems participants will be able to:
- Analyze core security and privacy challenges in agentic AI, from transparency gaps to data exposure
- Evaluate vulnerabilities in single-agent LLM deployments and apply targeted mitigation strategies
- Secure multi-agent LLM workflows against coordination failures and adversarial attacks
- Leverage privacy-enhancing technologies to protect sensitive data in autonomous AI pipelines