Agentic AI is increasingly taking over key tasks in modern platforms – from tool usage to automating complete workflows. This new autonomy brings enormous opportunities, but also opens up completely new attack surfaces.

In this session, we will show in a guided live demo how an Agentic AI system can be compromised step by step. Using practical scenarios, we make typical vulnerabilities visible: manipulated inputs, poisoned knowledge sources and abused integrations that can lead to unwanted actions and data leakage – and why classic AppSec reaches its limits here.

Speakers:

Nico Meisenzahl (COO, white duck)