How to leverage generative AI in cloud apps without putting user data at risk

This report provides a threat-modeling guide for securing multi-tenant cloud applications that use generative AI, applying the PEACH tenant isolation framework to assess risks and recommend controls. It details attack vectors (data poisoning, hallucination abuse, indirect/direct prompt injection), analyzes interface complexity across inference, customization, and feedback, and proposes architectural patterns (multi-tenant, single-tenant, hybrid) with practical isolation and hygiene recommendations to prevent cross-tenant data exposure and abuse.