Sovereign AI Data Preparation for Government Agencies
Ertas Data Suite gives government agencies an air-gapped, on-premise data preparation pipeline for building AI training datasets from sensitive government records — maintaining data sovereignty and satisfying federal security frameworks.
The Challenges You Face
Data Sovereignty Is a Mandate, Not a Preference
Government data — citizen records, law enforcement files, defense information, infrastructure data — must remain under government control at all times. FedRAMP, FISMA, CMMC, and agency-specific policies prohibit processing sensitive data on unauthorized external systems.
Classification Levels Restrict Tool Choices
Data at CUI, Secret, and Top Secret levels cannot be processed by most commercial AI tools. Even FedRAMP-authorized cloud services may not meet the requirements for certain data types, and the authorization process takes years.
Government Documents Have Unique Structures
Federal forms, regulatory filings, intelligence reports, and procurement documents follow government-specific formats, acronym conventions, and classification marking systems that commercial data tools do not understand.
Accountability and Traceability Are Constitutional Requirements
Government use of AI must be transparent, accountable, and auditable. Executive orders and OMB guidance require agencies to document how AI models are trained, what data they use, and how decisions are made — requirements that ad-hoc ML workflows cannot satisfy.
How Ertas Solves This
Ertas Data Suite is a native desktop application that operates in complete air-gap mode — no network connectivity, no telemetry, no external dependencies. Install it on a government workstation behind any classification boundary and prepare AI training datasets from the most sensitive government data without any risk of data exfiltration.
The five-module pipeline handles the full data preparation lifecycle. Ingest normalizes government document formats — PDFs, XML schemas, fixed-width text files, and database exports. Clean standardizes formatting, handles government-specific abbreviations, and removes irrelevant content. Label provides a structured interface for subject matter experts to annotate data. Augment generates controlled variations for balanced training. Export produces versioned datasets with complete provenance.
Every action is recorded in an immutable audit trail that documents who processed what data, when, and what transformations were applied — providing the accountability documentation that federal AI governance frameworks require.
Key Features for Government & Public Sector
Complete Air-Gap Operation
Data Suite functions with zero network connectivity. No DNS lookups, no update checks, no telemetry of any kind. The application is self-contained and fully operational on workstations with no network interface — suitable for SCIF and classified environments.
Government Document Format Support
The Ingest module handles government-standard formats including PDF/A, XML schemas (NIEM, etc.), fixed-width text exports from legacy systems, and structured data from federal databases. Custom parsers can accommodate agency-specific formats.
Federal Audit Trail Compliance
The immutable audit log captures every operation with the detail required by NIST AI RMF, OMB M-24-10, and EO 14110 guidance on AI accountability. Export audit records in formats compatible with agency ISSO documentation requirements.
Subject Matter Expert Labeling
Government analysts, intelligence professionals, and domain experts label data using the Label module's structured interface. Their institutional knowledge becomes encoded in training datasets without requiring data science skills.
Why It Works
- Data Suite's air-gapped architecture satisfies NIST SP 800-171 and CMMC Level 3 requirements for CUI processing without requiring additional security controls beyond the workstation's existing ATO.
- Government agencies have used Data Suite to prepare training datasets from controlled unclassified information without modifying their existing security architecture or obtaining new ATOs.
- The immutable audit trail provides the documentation required by OMB M-24-10 for agencies using AI in decision-making processes that affect individuals' rights or safety.
- Subject matter experts with no data science background have prepared high-quality labeled datasets using Data Suite's guided workflow, enabling AI projects that were previously blocked by the shortage of cleared ML engineers.
- Data Suite's deterministic pipeline ensures that identical inputs always produce identical outputs — a requirement for the reproducibility standards in the NIST AI Risk Management Framework.
Example Workflow
A federal agency wants to train a model that classifies incoming FOIA requests by complexity and routes them to the appropriate processing team. An analyst opens Ertas Data Suite on a workstation within the agency's CUI enclave, ingests 8,000 historical FOIA requests through the Ingest module, and runs the Clean module to normalize the varied submission formats.
Experienced FOIA officers use the Label module to classify each request by complexity tier, subject matter area, and typical response time. The Augment module generates variations of underrepresented request types. The Export module produces a versioned JSONL dataset with complete chain-of-custody documentation.
The dataset is transferred via approved media to the agency's training environment, where it produces a classification model that pre-screens incoming requests — routing complex cases to senior officers immediately and batching routine requests for efficient processing. Average initial routing time drops from days to minutes, with full audit documentation satisfying OMB reporting requirements.
Related Resources
Ship AI that runs on your users' devices.
Early bird pricing starts at $14.50/mo — locked in for life. Plans for builders and agencies.