State of Open Con 2025

Your models are running, your clusters are purring, and everything seems ready to sail smoothly across the vast seas of AI. Everything’s good, right? Attacks targeting GPUs, especially those aimed at poisoning AI models during training and inference, represent a growing frontier—much discussed but rarely explored.
In this hands-on talk, we’ll dive deep into how GPUs can be attacked and, more importantly, how to defend against these threats. You’ll discover best practices and learn how open-source tools you already know—like Falco, Cilium, and others—can protect your precious models. Get ready for an adventure into the open field of GPU security in AI. See it, Hack It, Sort It.