Deploying AI Systems Securely
Deployment guidance for organizations operating AI systems, with practical controls for the hosting environment and ongoing maintenance.
Published by National Security Agency and international partners
How to use it
Use this publication as a launch checklist when putting a model or AI application into an existing environment. It addresses the infrastructure hosting the system, how to protect model components, and how to run the system securely over time.
Give production access to named roles, isolate sensitive deployment assets, and set up monitoring that supports incident investigation. Document how to disable or roll back the AI component if it behaves unexpectedly.
- Access
- Public guidance