.svg){kind=icon}
Learn
Defense AI modernization is the Department of War effort to adopt artificial intelligence responsibly across missions, coordinated by the CDAO and shaped by initiatives like Task Force Lima, Replicator, and Responsible AI guidance.
Why it matters to your mission. Adopting AI across the force is less about any single model than about doing it responsibly, at scale, without losing control of data or decisions. A coordinated approach under CDAO guidance and initiatives like Task Force Lima and Replicator is what turns pilots into fielded capability. The outcome is AI that reaches real missions on a governed path, rather than demos that never scale.
Defense AI modernization is the Department of War effort to adopt artificial intelligence responsibly across missions, coordinated by the CDAO and shaped by initiatives like Task Force Lima, Replicator, and Responsible AI guidance.
Adoption succeeds only when mission data is AI-ready and the reasoning over that data is government-owned, traceable, and governable. A capable model trained elsewhere does not help a commander who cannot see the data behind an answer, cannot audit how it was produced, and cannot keep operating when a vendor relationship ends. The modernization initiatives below set the direction; the hard part is making fragmented systems of record usable without replacing them.
The scale of that fragmentation is documented. In 2025 the U.S. Government Accountability Office found the Department still moving information between command and control systems through manual "swivel chair" processes, with no Department-wide framework to guide investment or track progress toward integration (see GAO, GAO-25-106454, Defense Command and Control, 2025). Modernization that ignores that condition produces models with nothing trustworthy to reason over.
Defense AI modernization is the coordinated adoption of artificial intelligence across the Department of War so that AI reaches real missions on a governed, auditable path rather than staying trapped in pilots. It covers the policy, the data readiness, and the governance that let a model's output be trusted by a commander, not just demonstrated to one. The initiatives in the sections below are the parts of that effort most often cited by name.
The Chief Digital and Artificial Intelligence Office (CDAO) is the Department's senior organization for data, analytics, and AI. It sets the direction captured in the DoW 2023 Data, Analytics, and AI Adoption Strategy, which treats data as a strategic asset and argues that decision advantage comes from quality data and sound analytics, not from models alone. The CDAO's role is less to build one model for everyone than to make the conditions for trustworthy AI consistent across a Department whose data lives in many incompatible systems.
Task Force Lima was the CDAO-led effort to assess, synchronize, and responsibly employ generative AI across the Department. Its focus was figuring out where large language and generative models can genuinely help a mission, and where governance, traceability, and reliability have to be established before those models touch a decision. The task force was time-bounded; its lasting contribution is the principle that generative capability is adopted behind a governance line, not ahead of it.
Replicator is a Department initiative to field large numbers of attritable, autonomous systems at speed. As more of a mission involves autonomy, the governance and traceability of the reasoning behind those decisions moves from a nice-to-have to a core requirement. NIST frames human oversight as a risk-management measure for generative and autonomous AI, recommending that output accuracy and reliability be checked against known ground truth through methods that include human oversight in its Generative AI Profile (NIST AI 600-1, 2024). Fielding autonomy at scale does not remove the human; it raises the bar on being able to explain what the autonomy did and why.
The Responsible AI (RAI) Toolkit is the CDAO resource for applying the DoW's five principles of AI ethics - Responsible, Equitable, Traceable, Reliable, and Governable - across an AI system's lifecycle rather than as a one-time review. The point of a toolkit is that the principles become steps a program actually performs, with artifacts, instead of a slide a program nods at. Federal policy reinforces this: OMB Memorandum M-25-21 (2025) directs agencies to apply minimum practices to "high-impact" AI - uses whose output is a principal basis for decisions with a legal, material, binding, or significant effect on rights or safety - and to be able to discontinue a high-impact use that is not performing.
Pilots stall for reasons that have little to do with the model and everything to do with the conditions around it. The data a pilot ran on was curated for the demo and does not exist in that form in production. The output cannot be traced back to its sources, so an authorizing official cannot sign off. The capability was built inside a vendor's environment, so the Department cannot run, govern, or evolve it independently. And the integration work - connecting fragmented systems of record - was assumed away. Modernization is the discipline of fixing those conditions first, which is why the initiatives above lead with data readiness and governance rather than with models.
The initiatives are complementary, not competing. Each addresses a different part of getting AI to a mission responsibly, and each carries a readiness implication a program has to plan for.
| Initiative | Lead or source | What it targets | Readiness implication |
|---|---|---|---|
| Data, Analytics, and AI Adoption Strategy (2023) | CDAO / DoW | Data as a strategic asset | Mission data must be usable and trusted before models help |
| Task Force Lima | CDAO | Responsible generative AI | Governance and traceability come before fielding |
| Replicator | DoW | Attritable autonomous systems | Autonomy at scale raises the governability bar |
| Responsible AI Toolkit | CDAO | Five AI ethics principles in practice | Traceable and governable by design, across the lifecycle |
| OMB M-25-21 | OMB | Federal use of high-impact AI | Minimum practices and a plan to discontinue if performance drops |
Five initiatives, each a different part of fielding AI
The initiatives are complementary, not competing. Each addresses a different part of getting AI to a mission responsibly, and each carries a readiness implication a program has to plan for.
Pilots usually stall not because of the model but because production data is fragmented, output cannot be traced for sign-off, or the capability is locked inside a vendor environment.
Torch.AI builds government-owned reasoning infrastructure that complements existing systems of record rather than replacing them. It connects to mission data where it lives, makes that data AI-ready, and keeps the reasoning over it traceable and governable - so the governance these initiatives require is enforced in the infrastructure, not promised in a slide. This is the distinction between a system of record and a system of reason: the reasoning layer sits on top of the authoritative sources and preserves them, instead of migrating a mission onto a new proprietary platform.
Within that infrastructure, the public products carry the same intent into the field: NEXUS makes data meaningful through semantic representation and knowledge domains, HALO makes relationships understandable through graph-based fusion, and ORCUS provides a government-owned data-availability layer - with the reasoning kept doctrine-grounded, traceable, and auditable throughout. Whether a specific deployment is government-owned (GOTS) or commercial (COTS) depends on how the capability is purchased and installed; the point is that the government-owned model is available for missions where ownership and control decide the outcome. You can see how this is packaged on the Torch.AI software page.
The initiatives set the direction; the checklist below is what to actually require when evaluating a capability against it.
Evaluating a capability against these initiatives? Bring the seven requirements to a scoping call and we will walk each one against your environment: request a technical walkthrough.
Defense AI modernization is the coordinated adoption of artificial intelligence across the Department of War so that AI reaches real missions on a governed, auditable path, coordinated by the CDAO and shaped by initiatives like Task Force Lima, Replicator, and Responsible AI guidance.
The Chief Digital and Artificial Intelligence Office is the Department's senior organization for data, analytics, and AI, driving the DoW 2023 Data, Analytics, and AI Adoption Strategy, which treats data as a strategic asset and holds that decision advantage comes from quality data and analytics, not models alone.
Task Force Lima was the CDAO-led effort to assess, synchronize, and responsibly employ generative AI across the Department, focused on where generative models help a mission and where governance and traceability must be established first.
Replicator is a Department initiative to field large numbers of attritable, autonomous systems at speed, which raises the stakes on the governability and traceability of the reasoning behind autonomous decisions.
The Responsible AI Toolkit is the CDAO resource for applying the DoW's five AI ethics principles - Responsible, Equitable, Traceable, Reliable, and Governable - across an AI system's lifecycle rather than as a one-time review.
Pilots usually stall not because of the model but because the production data is fragmented, the output cannot be traced for sign-off, the capability is locked inside a vendor environment, or the integration work was assumed away - the conditions modernization exists to fix first.
They are complementary: the Adoption Strategy targets data readiness, Task Force Lima targets responsible generative AI, Replicator targets autonomy at scale, the RAI Toolkit operationalizes the five ethics principles, and OMB M-25-21 sets minimum practices for high-impact AI use.
Describe the problem.
Define the constraints. We will structure the session.
No demo.
No slide deck.
Real work.
Responsible defense AI starts with data readiness and traceable reasoning. We will scope it with your team. No slide promises. Your systems of record. Real work.
