Analysis in Gov AI reads from the same governed entity store as collection, so the graph, the timeline and the map show the same subjects without re-keying. Structured analytic techniques are built in, from competing hypotheses to key assumptions and challenge reviews, and AI assessments are saved as reviewable records with their sources. Findings move through review to classification-marked reports and sealed evidence bundles.
Explore relationships among people, organizations, accounts and places on an interactive graph with centrality measures, community detection, path finding and neighborhood isolation. An AI network assessment names the hubs, brokers and highest-risk actors for an analyst to review.
Swimlane timelines, density bands and a time scrubber establish sequence and bursts of activity. The signal lab detects anomalies and bursts across any collected series, projects them forward with prediction intervals and shows which signal tends to move first.
Hypothesis boards rank competing explanations by inconsistency and show which evidence is diagnostic. A key assumptions register arms each premise with falsifiers, challenge reviews let an independent colleague contest a product before release, and the scenario studio tests what-if assumptions on real baselines as labeled model output.
Compare writing style across accounts on phrasing, spelling, vocabulary, habits and rhythm, measured against a stated reference corpus and capped when a signal cannot be measured. Results support coordination and identity work as automated inferences, never as proof.
Structured assessments state a threat level, narratives, entities, claims and recommended actions, and are saved as records rather than chat. Findings carry an assertion type, a confidence band, evidence for and against, and citations, and move through review and publication with versions. Horus, the page-aware copilot, answers questions in the context of the page an analyst has open.
Produce classification-marked DOCX and PDF reports with charts, publish them to a library and distribute them by email to recipients cleared for the marking. Evidence bundles number and hash each exhibit, support redaction and a disclosure schedule, and seal the manifest so it can be verified later.
It produces structured intelligence products rather than open-ended chat. Each assessment states a threat level, the narratives in play, the entities involved and recommended actions, and is saved as a record that can be reviewed, compared and cited like any other product.
Every assertion keeps its type. Model output is labeled as an automated inference or a hypothesis, analyst conclusions as judgments and source material as statements, and model-inferred hypotheses are never scored as evidence. AI output is written back as a reviewable candidate, never as settled fact.
Yes. Reports carry classification markings, and distribution checks each recipient’s clearance against the marking. Markings support handling inside the platform; formal classification remains the responsibility of your government security authority.
Intelligence products may be challenged later, in oversight or in court. A tamper-evident record of who collected, analyzed, approved and released each item shows how a conclusion was reached and that the record has not been altered.
See analysis and reporting working on your own sources, in a secure environment with clearance verification.