Build a Platform and Watch It Burn — Michael Forrester, Accenture & Whitney Lee, Datadog
Unprotected AI agents on internal platforms can hack Kubernetes clusters via prompt injection
“even if your agent is not protected, it can cause reputational damage in addition to financial and data-related losses”
Accenture and Datadog engineers ran a live workshop demonstrating how an unguarded LLM-powered 'burrito bot' deployed on an internal developer platform (IDP) can be manipulated to compromise a Kubernetes cluster. The talk illustrates real-world risks of shipping agentic systems without safety fuses, using a deliberately vulnerable 34-component IDP stack. It's a practical red-team demo rather than a research finding, but its engineering audience and live exploitation angle make it a useful signal for teams building agentic platforms.