This conference talk explores how to build a zero-trust security architecture for AI workloads using cloud native patterns, specifically focusing on Large Language Model (LLM) applications. Learn essential strategies for implementing AI Gateways with robust authentication, authorization, and audit logging capabilities. Discover methods to maintain compliance and governance requirements while securing model artifacts, implementing runtime security, and defending against prompt injection attacks. The presenters, Rohit Ghumare from Taikun and Joinal Ahmed from NTG, provide practical guidance for organizations looking to protect sensitive information and ensure smooth operations as they increasingly rely on LLM applications for critical business functions.