More value, less risk: How to implement generative AI across the organization securely and responsibly

This Microsoft Security strategy paper provides guidance for securely and responsibly deploying generative AI, focusing on five areas: data security, managing hallucinations and overreliance, defending against threat actors (including jailbreaks and malicious instructions), addressing biases, and ensuring legal and regulatory compliance. It recommends grounding outputs on trusted data, AI red teaming, strong authentication and insider risk programs, ethics committees, and alignment with frameworks like the NIST AI RMF, using a map–measure–manage approach to balance innovation with risk.