Secure Architecture Review for Generative AI Deployments

A security architect reviewing a proposed internal or customer-facing generative AI application needs to systematically identify prompt injection risks, data leakage vectors, model access control weaknesses, and unsafe output handling patterns before the application goes to production. Most existing secure design frameworks predate LLM deployment patterns.