Executive Summary
Artificial intelligence is being integrated into defense operations, acquisition support, and mission systems at an accelerating pace. The DoD has published multiple foundational documents governing this integration — the DoD AI Strategy, the Responsible AI guidelines, and use-case-specific policies from CJADC2 and the various service branches. This article reviews what the current policy framework actually permits, where AI is being piloted, and what the compliance and oversight guardrails look like for commercial AI companies seeking defense market entry.
The Policy Foundation
The DoD AI Strategy (2018, updated 2023) establishes five core principles for responsible AI use: responsible, equitable, traceable, reliable, and governable. These are not marketing language — they are evaluation criteria that acquisition offices use when assessing AI-enabled capabilities. A commercial AI product that cannot demonstrate traceability (explainable outputs) or governability (human oversight and correction mechanisms) faces significant evaluation challenges in defense procurement.
The DoD Directive 3000.09 on autonomous weapons establishes requirements for human judgment in lethal force decisions. While most commercial AI companies are not building lethal autonomous systems, the oversight architecture defined in 3000.09 has influenced how DoD evaluates AI governance across all categories.
Where AI Is Being Authorized
Current authorized or piloted AI use cases in defense acquisition and operations include:
- Logistics and predictive maintenance: AI models analyzing equipment sensor data to predict maintenance needs before failure. Army Materiel Command has multiple programs in this space.
- Contract writing support: AI-assisted draft generation for standard contract types, with human review and approval required before issuance.
- Intelligence analysis support: AI-assisted processing of large volume structured data sets, with analyst verification of outputs before inclusion in intelligence products.
- Cybersecurity threat detection: AI-enabled anomaly detection in network traffic, with security analyst review of flagged events.
- Administrative process automation: Routing, scheduling, and workflow automation in non-sensitive operational contexts.
What Is Not Authorized (Or Constrained)
Lethal autonomous weapons require human judgment at the point of lethal force application — this is not negotiable under current DoD policy. High-stakes personnel decisions (promotion, assignment, separation) have significant oversight requirements and legal constraints. AI systems that produce outputs that cannot be explained or audited face substantial barriers in formal acquisition programs.
The Commercial AI Gap
Most commercial AI companies entering the defense market underestimate the governance and traceability requirements. A system that performs well on commercial benchmarks may fail defense evaluation because it cannot provide: audit logs of decision inputs and outputs, explanation of how specific outputs were reached, mechanisms for human override at defined decision points, and documentation of training data sources and validation methodology.
These are not optional features — they are threshold requirements in responsible AI policy.
Operational AI: A Specific Consideration
AI systems that operate in or adjacent to mission-critical functions (not just administrative processes) face additional scrutiny under JADC2 integration requirements. Cross-domain solutions, data handling, and interoperability with existing DoD systems create compliance surface area that commercial AI companies often do not anticipate until late in the acquisition process.
Recommendations for Commercial AI Companies
Build traceability into the architecture from day one, not as a compliance retrofit. Document training data provenance. Implement human-in-the-loop override at all decision points where the policy framework requires it. Engage with the Responsible AI for DoD principles (freely available from CDAO) early in product development, not after DoD market entry.
SOSRA's Analytical Position
SOSRA monitors DoD AI policy developments and provides analysis to help organizations understand the compliance landscape before they engage with defense acquisition. We do not endorse specific AI products, claim government partnerships, or guarantee acquisition outcomes. Our role is analytical and advisory.
