AI705 Gap Analysis for Frontier AI Facilities
Technical guidance for applying ICD 705 and IC Tech Spec requirements against weight theft, secret theft, and sabotage.
About this project
A shared reference for applying ICD 705 to frontier AI facilities.
Security Level 5 is a security posture for AI systems that could plausibly thwart top-priority operations by the world's most cyber-capable institutions. AI705 focuses on one long-lead part of that posture: how classified-facility standards should be interpreted for frontier AI datacenters and model-weight operations.
ICD 705 and the IC Tech Spec were written for Sensitive Compartmented Information Facilities. Frontier AI labs, colocations, and dedicated datacenter operators are now asking how that body of practice should apply to non-classified but extremely high-value assets, facility secrets, and sabotage-sensitive infrastructure.
AI705 is a requirement-by-requirement gap analysis. It identifies which facility requirements apply as written, which need modification, which do not apply, and where new controls or measurement campaigns are needed.
Guidance Draft
Read the pilot AI705 guidance structure and current gap-analysis rows.
2Threat Vectors
Track weight theft, secret theft, sabotage, and the technical channels that support evidence review.
3References
Source standards, supporting references, and publication citations.
4About
SL5 Task Force context, project status, and contribution path.
Theory of change
Concrete guidance reduces improvisation at the facility boundary.
If facilities teams, colo providers, and accrediting officials can evaluate frontier AI environments against a shared AI705 reference, they can design and review physically secure AI infrastructure more consistently. That shared baseline is especially valuable where the original SCIF framing does not directly address accelerator racks, liquid cooling, shared datacenter services, model-serving operations, sabotage paths, or AI-specific side-channel concerns.