RDIV / ANSWER AI-006 · SOVEREIGN INTELLIGENCE
How Should Governments Procure AI Systems?
A source-led framework for evaluating AI procurement through data rights, auditability, continuity, portability, human authority, and credible exit options.
Fix the object first.
AI procurement converts governance principles into enforceable technical and contractual conditions. The exact legal requirements depend on the agency and jurisdiction; this page therefore separates current U.S. federal guidance from the broader RDIV governance checklist.
The visible label is not the whole system.
Procurement is upstream governance. Once a system becomes embedded in data flows and workflows, renegotiating audit access, portability, model-change notice, or exit rights can be expensive or impossible.
Trace the burden.
- OMB M-25-22 is current federal guidance on AI acquisition and explicitly emphasizes competition, data portability, interoperability, and avoiding costly single-vendor dependencies.
- OMB M-25-21 governs broader federal AI use and governance.
- NIST and GAO provide complementary risk-management and accountability frameworks.
- Agency-specific statutes, appropriations, acquisition rules, security requirements, and mission constraints can add obligations not summarized here.
- Current federal guidance can change and must be reverified before relying on this page as a policy reference.
Where the work adds something.
RDIV’s procurement lens adds a command checklist: before contract award, determine whether the agency can inspect, audit, contest, suspend, replace, continue, and exit. These are governance recommendations unless a cited authority makes them binding.