This NIST draft examines security threats facing AI data centers for federal IT leaders, cybersecurity teams, and infrastructure operators. It compares AI computing environments with high-performance computing systems and presents a four-zone architecture covering computing, data storage, access, and management.
The analysis addresses risks involving AI model theft, data poisoning, prompt injection, shared infrastructure, firmware vulnerabilities, and supply chains. It recommends stronger gateway protection, continuous monitoring, zero trust, confidential computing, human oversight, and fault-tolerant operations. The publication focuses on the computing environment rather than physical facilities or operational technology, providing a foundation for future AI data center security guidelines.

| Format: |
|
| Topics: | |
| Website: | Visit Publisher Website |
| Author(s): |
|
| Publisher: | National Institute of Standards and Technology (NIST) |
| Published: | July 27, 2026 |
| License: | Public Domain |
| Tags: | AI Data Center SecurityArtificial IntelligenceHigh-Performance ComputingnistNIST SP 800-239 |