AI systems are being deployed faster than most organizations can assess them. Agentic AI (systems that autonomously plan, act, and interact with other systems) introduces a fundamentally new attack surface that traditional threat modelling approaches were not designed to handle.

This full-day workshop is built around doing, not just listening. Participants work through real AI system architectures, analyse actual failure cases, and apply threat modelling techniques hands-on. The mix of structured input and interactive exercises is designed for security architects, risk managers, and assessors who need to translate AI security principles into practice.

The Day Covers:

• Architecture of LLM-backed and agentic AI systems: what makes them different

• Examples of sovereign AI architectures

• Real-world agentic AI failures: what went wrong and why

• Common vulnerabilities found in AI applications by red teams

• Threat modelling techniques adapted for AI systems

• Risk and control frameworks for AI

• Allocating security responsibilities across parties: the OLRA model

• Securing agentic coding environments

• Key elements of the ISO 42001 AI management standard

Relevant for professionals pursuing TAISE or CCZT certifications.

Bonus: Course materials and supplemental materials will be provided digitally. There will also be two followup Q&A calls in the weeks after the workshop.

Your instructor: Dr. Peter van Eijk

One of the world’s most experienced cloud trainers: Technology for decision makers. Developing and delivering training and coaching programs for Cloud, AI and Security.