AI705 is being developed through expert review across facility security, AI infrastructure, and standards translation.

AI705 depends on review from communities that do not always share a common vocabulary: SCIF and facility-security practitioners, frontier AI infrastructure teams, datacenter operators, colocation providers, accrediting officials, side-channel researchers, OSCAL and compliance-tooling specialists, and policy reviewers.

We are especially grateful to reviewers willing to test whether this project’s central decomposition - weight theft, secret theft, and sabotage, with stored/training/inference categories for weight theft - matches the operational decisions facilities teams actually need to make.