This workshop goes beyond classic prompt injection and explores the broader threat landscape of prompt attacks in modern AI systems. We break down real-world agent architectures and examine how prompt attacks propagate through tools, memory, retrieval systems, and external integrations. Attendees will additionally have the option to work in an agent lab environment to actively exploit the systems discussed, observing how attacks move across components and trigger unintended behavior. The goal is to build a realistic understanding of how prompt attacks affect production-grade AI systems.
To register to this workshop, you can click here.
Entrance to this side event is permitted only with a valid ticket to Insomni’hack 2026!
