The State of AI in DoD: Trends and Opportunities for 2026

Artificial intelligence has moved from the periphery of Department of War strategy to its center. The establishment of the Chief Digital and Artificial Intelligence Office (CDAO) and the acceleration of Joint All-Domain Command and Control (JADC2) point in one direction. So does a growing portfolio of AI-enabled programs across every military service. Together, these signals reflect an institutional commitment to AI that is unprecedented in scope and urgency. Yet for all this momentum, the path from AI ambition to deployed AI capability remains fraught with technical, organizational, and regulatory challenges. This whitepaper examines the current state of AI in the Department of War and identifies the trends shaping its trajectory in 2026. In addition, it highlights the opportunities for defense contractors positioned to deliver real-world AI capabilities.

The CDAO and the Centralization of AI Strategy

The creation of the Chief Digital and Artificial Intelligence Office in 2022 represented a decisive shift in how the Department of War approaches AI governance. The office consolidated the Joint Artificial Intelligence Center (JAIC), Defense Digital Service, Advana, and the Chief Data Officer under a single organization. The goal was to eliminate the fragmentation that had hindered AI adoption across the enterprise.

In 2026, the CDAO’s influence continues to expand. The office has established data and AI governance frameworks that provide common standards for AI development, testing, and deployment across the services and combatant commands. Moreover, its focus on data readiness ensures that the Department of War’s vast data holdings are discoverable, accessible, and usable for AI applications. As a result, this focus has driven significant investment in data cataloging, metadata management, and data sharing infrastructure.

The CDAO has also championed the concept of AI as a shared service. It promotes reusable AI components, common development platforms, and centralized model repositories that reduce duplication and accelerate deployment. For defense contractors, this centralization creates both opportunities and requirements. Contractors can build components that serve multiple programs. However, they must also align with CDAO standards, interfaces, and governance processes.

JADC2: AI as the Connective Tissue of Multi-Domain Operations

The Joint All-Domain Command and Control vision connects sensors, shooters, and decision-makers across air, land, sea, space, and cyberspace into a unified information network. It is arguably the most ambitious technology initiative in Department of War history. AI is fundamental to JADC2 — not as an add-on, but as the enabling technology that makes multi-domain integration feasible at the speed of modern warfare.

Without AI, the volume of data produced by sensors across all domains would overwhelm human decision-makers. AI enables automated sensor fusion, threat identification, course-of-action development, and decision support at machine speed. For example, computer vision systems process full-motion video and satellite imagery. Natural language processing systems analyze signals intelligence and open-source reporting. Predictive models anticipate adversary actions and logistics requirements. Together, these AI capabilities transform raw multi-domain data into the situational awareness and decision advantage that JADC2 promises.

Each military service is pursuing its own JADC2 contribution — the Army’s Project Convergence, the Air Force’s Advanced Battle Management System (ABMS), and the Navy’s Project Overmatch. As a result, a broad landscape of opportunities exists for AI development, integration, and deployment. Contractors who can deliver AI systems across these service-specific architectures, while adhering to joint interoperability standards, will be positioned for sustained program success.

AI Adoption Challenges: The Gap Between Ambition and Deployment

Despite significant investment and strategic emphasis, the Department of War’s AI adoption continues to face structural challenges that slow the transition from prototype to production.

Data readiness remains the primary bottleneck. AI systems require large volumes of high-quality, labeled data for training. However, the Department of War’s data landscape is fragmented across classification levels, organizational boundaries, and legacy systems. Data often sits locked in stovepipes, stored in incompatible formats, or lacking the metadata necessary for AI consumption. The CDAO’s data governance initiatives are addressing these issues, but progress is measured in years, not months.

Integration with legacy systems presents persistent challenges. Many of the operational systems that AI must interface with — command and control platforms, intelligence databases, logistics systems — date back decades. Their designers never considered AI integration. Connecting modern AI capabilities to these systems requires middleware, API development, and careful interface engineering. In fact, this work is often more complex and time-consuming than the AI development itself.

The “valley of death” between successful AI prototypes and fielded AI capabilities continues to claim promising projects. A model that achieves impressive accuracy in a development environment must still survive security accreditation, operational testing, user training, and sustainment planning. Only then does it deliver value to the warfighter. Many AI initiatives lack the engineering and programmatic support to navigate this transition. As a result, impressive demonstrations never reach operational users.

Cultural resistance and organizational inertia also slow adoption. Decision-makers accustomed to traditional analytical processes may be reluctant to trust AI-generated insights. Similarly, operators may resist changes to established workflows. Therefore, overcoming this resistance requires not just better technology but sustained engagement, training, and demonstration of AI value in operationally relevant contexts.

Responsible AI: Ethics and Governance in Defense Applications

The Department of War has adopted a responsible AI framework built around five principles: equitable, traceable, reliable, governable, and responsible. These principles are not abstract aspirations. Instead, they carry concrete implications for how AI systems are designed, developed, tested, and deployed.

Traceability requires that AI decision processes be auditable and explainable. In defense applications, an AI system may recommend a course of action or identify a target. When it does, analysts and commanders must be able to understand the basis for that recommendation. Black-box models that produce accurate results without explanation face significant adoption barriers in operational contexts where accountability is paramount.

Reliability demands rigorous testing across the full range of operational conditions. This includes adversarial scenarios where opponents may attempt to deceive or manipulate AI systems. Moreover, testing for bias, robustness, and edge-case behavior is essential. Teams must integrate this testing into the development lifecycle rather than treat it as a final quality check.

Governability requires clear human oversight mechanisms. AI systems in defense applications must have well-defined boundaries of autonomous action and clear procedures for human intervention. The level of human oversight should match the consequences of the AI’s decisions — higher stakes demand more human involvement.

For defense contractors, responsible AI is not optional. It is a requirement that shapes technical architecture (explainability layers, audit logging, bias detection), development processes (diverse training data, adversarial testing), and documentation (risk assessments, ethical reviews). Therefore, contractors who build responsible AI practices into their engineering culture will navigate the governance landscape more effectively than those who treat it as a compliance exercise.

ATO for AI: The Accreditation Challenge

The Authority to Operate process is the security accreditation framework that governs all IT systems in the Department of War. Its designers built it for traditional software systems, so it fits AI imperfectly. AI systems introduce novel security considerations that existing ATO frameworks are still evolving to address.

Model integrity means ensuring that no one has tampered with a deployed model or poisoned it through adversarial training data. This concern has no direct analog in traditional software security. Model drift is the degradation of model performance over time as the operational data distribution diverges from training data. It requires continuous monitoring capabilities that traditional ATO processes do not contemplate. In addition, the use of third-party models and pre-trained weights introduces supply chain risk that teams must assess and mitigate.

The Risk Management Framework (RMF) is being adapted to address these AI-specific concerns, but the process is ongoing. In the meantime, defense contractors must work closely with authorizing officials to develop ATO packages that credibly address AI-specific risks. This requires security engineering expertise that understands both the RMF process and the unique characteristics of AI systems. That combination remains rare in the defense workforce.

The Talent Gap: Defense AI’s Persistent Challenge

The Department of War competes for AI talent against technology companies offering significantly higher compensation, more flexible work environments, and fewer security constraints. In addition, clearing AI engineers and data scientists adds months of delay and eliminates candidates unwilling to submit to the investigation process. The result is a persistent talent gap that constrains AI adoption across the defense enterprise.

Addressing this gap requires creative approaches from both government and industry. Defense contractors can compete by offering meaningful mission impact: the opportunity to work on problems of genuine national significance. They can pair that impact with competitive compensation, professional development, and quality of life. Geographic flexibility also helps. For example, positions in lower-cost-of-living areas like Augusta, Georgia, help compensation packages stretch further. They also offer engineers a lifestyle that Silicon Valley cannot match.

Upskilling existing defense IT professionals in AI and machine learning is another critical strategy. Software engineers and data analysts with domain expertise and active clearances can learn AI techniques quickly. That path is far more efficient than clearing AI researchers and orienting them to defense missions. Investment in training and development is not just a benefit — it is a workforce strategy.

Zapata Technology: Delivering Deployed AI for Defense

At Zapata Technology, we have moved beyond AI theory and into AI practice for the Department of War. Our products and services reflect the hard-won lessons of building, deploying, and operating AI systems in real defense environments.

CASCADE, our AI/ML framework, provides the development and deployment infrastructure that defense AI programs need. It spans data preparation and model training through deployment on classified networks and edge devices. Furthermore, CASCADE’s architecture addresses the data readiness, security accreditation, and operational deployment challenges that stall so many defense AI initiatives.

LIGHTNER, our object recognition tool, demonstrates what deployed defense AI looks like in practice. Optimized models run on tactical hardware in disconnected environments and deliver real-time intelligence to warfighters without cloud dependencies. In short, LIGHTNER embodies the edge AI capabilities that JADC2 and tactical modernization demand.

Our AI and machine learning services extend beyond our products to encompass the full spectrum of defense AI needs. These include data engineering, model development, MLOps, security accreditation support, and operational sustainment. We understand that deploying AI in defense is not just a technology challenge. Rather, it is an integration challenge that spans technology, security, operations, and people.

Looking Ahead: Opportunities in Defense AI for 2026 and Beyond

The trajectory of AI in the Department of War points toward several opportunities for defense contractors prepared to deliver:

Edge AI and tactical deployment will continue to grow as JADC2 matures and services field AI-enabled systems closer to the point of need. Contractors who can optimize models for constrained hardware, operate in disconnected environments, and meet the ruggedization demands of tactical platforms will find expanding opportunities.

Data engineering and data readiness will remain foundational. As the CDAO drives enterprise data governance, demand will grow for data engineers who can build pipelines, manage metadata, and ensure data quality in classified environments. After all, AI cannot outrun its data foundation.

AI security and accreditation will emerge as a distinct discipline. As AI systems move from prototypes to production, demand will increase for engineers who can navigate the ATO process for AI. These engineers must also address model-specific security risks and implement continuous monitoring.

Human-AI teaming will define the next phase of operational AI. AI must move beyond a back-office analytical tool and become an integrated partner in operational decision-making. That shift will require new interface designs, trust frameworks, and operational concepts that blend human judgment with machine speed.

The Department of War’s AI journey is still in its early chapters, but the direction is clear and the investment is real. Some contractors have the technical depth, security expertise, and operational understanding to deliver AI that works in the real world, not just in demonstrations. For them, the opportunities ahead are substantial and enduring.