Organizations are rapidly deploying AI, yet security teams are uncertain about assessing these new surfaces. Secure Ideas' Kevin Tackett pulls from real-world pentesting engagements to demonstrate how AIsystems are vulnerable to attack. Kevin will walk through various AI implementations, including LLMs, ML pipelines, and AI-integrated applications. Using stories, he will cover the risks these systems introduce, from prompt injection to API abuse. Each scenario will be grounded in actual tests, illustrating how misconfigurations or overlooked trust boundaries led to compromise.