Threat modeling in AI is often an afterthought, yet ignoring risk during ideation creates blind spots that cascade into architecture, training, and trust. This session explores how to design secure AI systems before a single line of code is written. Drawing from real-world case studies, Rakia Finley presents frameworks for aligning threat modeling to model ethics, data governance, and long-term resilience from day zero.