Securing the Agentic Enterprise

A practitioner course on securing AI agents – systems that act, not just answer. Learn how agents work and break, how prompt injection and machine identity sprawl create real incidents, and how to build layered controls, detection and a 90-day programme. Based on ICCSO's Securing the Agentic Enterprise practitioner guide.

Who this course is for: SOC analysts, security engineers and architects, IAM teams, IT managers, CISOs and security leaders, including small one- and two-person security functions. Basic cyber security knowledge is assumed. No machine-learning expertise is needed. ICCSO's AI Security and Governance course is useful but not required.

Self-paced online learning. Earn a Certificate of Completion on passing.

What you will learn

• Explain how AI agents work – goal, model, tools, memory and environment – and where the identity boundary sits
• Threat-model agentic systems using the eight agentic attack surfaces and the OWASP Top 10 for Agentic Applications
• Reduce the impact of prompt injection through blast-radius design, including avoiding the lethal trifecta
• Govern machine identities for agents using least privilege, short-lived credentials and per-agent identities
• Discover and govern shadow AI, and secure RAG pipelines and agent memory
• Apply a six-layer defence-in-depth framework, agentic detection and response, and a 90-day roadmap

Course modules

• The agentic shift
• Prompt injection in agentic systems
• Machine identities at scale
• Shadow AI, RAG pipelines and memory
• Building the agentic security programme