We are looking forward to the Cloud Native Austria Days and expect great sessions as well as networking with the cloud, open source community.

“Agentic AI Under Attack: Live Demos of Exploits Through Autonomy and Trust” is the title of the session by our COO Nico Meisenzahl.

AI agents make decisions, invoke tools, and trust data without human review. However, this autonomy creates a new and largely misunderstood attack surface. In this demo-based talk, we show how agent-based AI systems can be hijacked without code exploits. Only text, tools and trust are used.

Using live demos, we explore three real-world classes of vulnerabilities from the OWASP Top 10 for AI:

– Indirect prompt injection, where untrusted content silently manipulates
the agent’s decisions- Tool/MCP poisoning, where malicious tool descriptions hijack the agent’s behavior and expose
the entire context- RAG poisoning, where internal knowledge leads to ongoing data exfiltration

No slides. No theory. Demo only! With practical DevSecOps lessons on why classic security controls fail as soon as AI agents act on your behalf.