OHaaS is purpose-built to operationalize Secretary Hegseth's AI Acceleration Strategy — delivering hardened, classified-ready AI agent deployments from IL5 through Top Secret, at the speed the mission demands.
In January 2026, Secretary of War Pete Hegseth launched the AI Acceleration Strategy — directing the Department of War to become an "AI-first" warfighting force by eliminating bureaucratic barriers and unleashing American AI innovation at mission speed.
The strategy centers on four lines of effort and seven pace-setting projects designed to fast-track AI-enabled capabilities across warfighting, intelligence, and enterprise operations.
Empower warfighters and innovators to rapidly discover, test, and scale AI capabilities in real conditions — not endless pilot programs.
Cut red tape that delays AI adoption. Small, accountable teams with authority to deploy at the speed of relevance.
Concentrate resources on U.S. asymmetric advantages — frontier AI models, compute infrastructure, and elite talent pipelines.
Deliver real AI capabilities to the warfighter now. Build quickly, deploy what works, fix what doesn't — no more PowerPoint wars.
OHaaS was designed for exactly this moment. Every capability maps directly to the AI Acceleration Strategy's lines of effort and pace-setting projects.
| Strategy Requirement | OHaaS Capability | How It Works |
|---|---|---|
| GenAI.mil Integration Pace-Setter |
Ready-to-deploy configurations for DoD AI infrastructure | Pre-configured to work seamlessly with existing GenAI.mil and GovCloud AI services. Deploy agents that integrate with DoD-approved AI endpoints out of the box. |
| AI-First at Every Classification Line of Effort |
IL5 through Top Secret deployment options with full network isolation | Deploy AI agents at any classification level without cross-contamination risk. Each classification operates as an isolated environment with appropriate security controls. |
| Eliminate Bureaucratic Blockers Line of Effort |
Deploy new AI agents in minutes, not months | Streamlined provisioning eliminates lengthy approval cycles. Teams can spin up secure, compliant AI agent instances on-demand without waiting for separate ATO processes. |
| DoD CAC/PIV + YubiKey Auth Security |
Native support for DoD smart card authentication | Warfighters access AI agents using their existing CAC/PIV credentials. Multi-factor authentication options support diverse user populations while maintaining zero-trust security. |
| Azure Gov / GCC High Infrastructure |
Built for GovCloud and classified environments | Designed from the ground up for government cloud infrastructure. Zero commercial cloud dependencies. Supports Secret and Top Secret classification levels. |
| Swarm Forge / Open Arsenal Pace-Setter |
Rapid experimentation with full security controls | Enable teams to test novel AI-enabled capabilities in isolated environments. Scale successful experiments while maintaining complete visibility and control over network access. |
| FIPS 140-3 Compliance Mandate |
Chainguard FIPS-validated base images — zero CVE, zero Docker Hub dependencies | Chainguard STIG-compliant and FIPS 140-3 validated base images. Cryptographic modules validated. Zero CVE policy enforced. |
| Content Security Security |
DLP scanning, prompt injection detection, malicious code detection | Outbound DLP catches PII/credentials before exfil. Inbound prompt injection detection blocks manipulation attempts. YARA-based code scanning stops reverse shells, cryptominers, and container escapes. |
| Speed of Relevance Core Principle |
Rolling upgrades across all tenants in a single command | ohaas upgrade 2026.x.x updates every tenant simultaneously. Canary rollouts, automatic rollback, zero-downtime deployments. |
OHaaS inspects every message in and out of your AI agent — detecting threats, stopping data leaks, and blocking malicious activity in real time.
Monitors every response from your AI model for attempts to hijack agent behavior — including jailbreak attempts, instruction overrides, and commands designed to exfiltrate data.
Scans every outbound AI request for sensitive data — PII, credentials, financial information, controlled unclassified information (CUI), and classified markings — before it leaves your environment.
Analyzes every command your AI agent executes for signs of malicious intent — unauthorized network connections, attempts to escalate privileges, or patterns associated with known attack tools.
Your warfighters and acquisition professionals don't need to be AI experts. These skills give them mission-ready capabilities from day one.
All skills run on FIPS-validated images, fully air-gapped compatible — no data leaves the enclave.
OHaaS supports three air-gapped classification presets, each with its own isolated egress rules, cloud endpoints, and network boundaries. No cross-contamination between levels.
NIPRNet
DoD-approved AI endpoints
Government cloud services
Identity & authentication
Certificate validation
SIPRNet
DoD SECRET network endpoints
Classified cloud services
Identity & authentication
Certificate validation
+ Optional isolated cloud
JWICs / NSANet
Intelligence Community endpoints
Top Secret cloud services
Classified identity providers
Certificate validation
+ Optional isolated cloud
Each level is a standalone air-gapped configuration — not cumulative. Minimal shared services for AI and certificate validation only.
The AI Acceleration Strategy defines seven pace-setting projects. OHaaS provides the hardened infrastructure layer for deploying AI agents across all of them.
Competitive AI experimentation needs isolated, disposable environments. OHaaS multi-tenancy lets elite units and innovators spin up secure AI agent sandboxes in minutes — test, learn, kill, repeat.
TechINT-to-capability in hours requires AI agents with controlled egress to classified intelligence systems. OHaaS egress rules ensure agents can reach only the endpoints they need — nothing more.
Department-wide generative AI access at IL5 and above. OHaaS routes all AI traffic through Ask Sage's GovCloud endpoints (*.asksage.mil, *.genai.army.mil) — the same platform powering GenAI.mil.
Top AI talent expects modern tooling. OHaaS gives each operator their own hardened AI agent environment — not a shared VM from 2015. Attract and retain the best by giving them the best.
AI agents with tool access need guardrails. ClawBands middleware enforces paranoid-mode security policies — approval gates, command restrictions, and real-time monitoring at every classification level.
When an AI agent behaves unexpectedly, kill its network access in seconds — not hours. OHaaS quarantine blocks all egress with a single command while preserving state for forensic analysis.
OHaaS is built for the DoW's AI-first future. Get started with a classified briefing.