Can Octostar be accredited for classified networks?

Octostar is designed to run inside an agency's own accreditation boundary. It deploys on-premise or air-gapped, has no dependency on external services, performs AI processing locally and provides the audit logging and access controls that accreditation processes require. Accreditation itself is carried out by the agency under its national scheme.

Which intelligence disciplines does it support?

The platform is discipline-agnostic. It has been applied to SIGINT, OSINT, HUMINT, FININT, ADINT and cyber data. Each source is mapped onto the shared ontology, so analysis works the same way regardless of where the data came from.

Do the AI models require a connection to an external provider?

No. Transcription, translation, image and video understanding, entity extraction and the generative assistant all run on local models inside your deployment. The Octobox appliance demonstrates this in a fully disconnected form.

How does Octostar compare with Palantir Gotham?

Both are all-source analysis platforms built around a knowledge graph. Octostar differs in being European-built, deployable and extendable independently of the vendor through an open API and APPs, and available on an appliance for air-gapped use. Agencies evaluating alternatives to US platforms typically start with an Octobox proof of concept on their own data.

Can partner agencies share a deployment?

Yes. Compartmented access control, per-record permissions and shared workspaces allow several units or partner agencies to work on one deployment while each sees only what it is cleared to see.

National Security

Sovereign Intelligence Analysis for National Security

Fuse SIGINT, OSINT, HUMINT and cyber data into one graph, at petabyte scale, on infrastructure that your agency owns and can evolve.

National security work depends on joining intelligence that arrives in different forms, from different collection disciplines, at different classification levels. Octostar is an all-source intelligence analysis platform that maps every source onto a shared ontology, so intercepts, open sources, human reporting and cyber telemetry resolve into the same entities and the same graph.

The platform is European-built and designed for sovereignty: it runs on-premise, in a national private cloud, or fully air-gapped, with all AI processing performed locally. Agencies keep control of the code they extend, the models they run and the data they hold. Intelligence Online has described Octostar as one of only two European alternatives to Palantir, and a collaboration with BAE Systems was announced in March 2026.

The platform overview covers the architecture in more depth; the sovereign intelligence page sets out the case for a European platform.

The challenge

Why all-source analysis is hard

Collection outpaces analysis

Intercept volumes, open-source feeds and cyber telemetry grow faster than analyst headcount. Without automation, most of what is collected is never read.

Disciplines do not share a data model

SIGINT, OSINT and HUMINT tooling each describe the same person, number or organisation differently. Fusion happens by hand, late, and is rarely repeatable.

Sovereignty is non-negotiable

Classified material cannot depend on foreign-hosted platforms or on a vendor's decision to keep supporting a market. Agencies need a system they can run and extend themselves.

AI has to be trusted before it is used

Models that cannot show their sources, that cannot be run offline, or whose behaviour cannot be audited will not be cleared for use on sensitive material.

Capabilities

Built for all-source fusion

Core capabilities of the platform as applied to intelligence work.

SIGINT processing at scale

APPs for signals processing turn intercept product into transcribed, translated, entity-tagged and searchable records, linked automatically to the selectors and subjects they concern.

OSINT on demand

Semantic drivers pull open-source and premium third-party data into an investigation as it is needed, mapped straight onto the ontology rather than landed as raw files.

Virtual knowledge graph

Sources stay where they are and are viewed as one interconnected graph. Analysts pivot from a selector to a person to an organisation to a location without changing tools.

Alerting and anomaly detection

Rules, watchlists and models run continuously over new collection and raise alerts on matches, pattern changes and anomalies, with the evidence attached.

AI assistant with provenance

Generative AI summarises reporting, answers questions and drafts assessments over the graph. Every output cites its sources and is logged, and models run inside your perimeter.

Compartmented access

Role-based and attribute-based access control, per-record permissions and full audit logging support need-to-know handling across teams and partner agencies.

Collaboration across units

Shared workspaces, real-time co-working on charts and structured reporting let collection, analysis and operations teams work on the same picture.

Air-gapped and on-premise

Kubernetes deployment on your hardware or national cloud; the Octobox appliance for disconnected sites. No component requires an external service.

Deployment

Deploying in a classified environment

Step 1
Evaluate on an Octobox
A self-contained appliance with the full platform and local AI. Bring it into a secure room, load representative data and evaluate against real analytic questions within a day.
Step 2
Map your sources
Collection systems and repositories are mapped onto the ontology through semantic drivers. Federate what must stay in place; ingest what needs sub-second search.
Step 3
Run it as your own system
Deploy on Kubernetes inside your accreditation boundary. Your engineers extend it through the open API and APPs, with source-level access where your programme requires it.

Why Octostar

Why intelligence agencies choose Octostar

Sovereign: European-built, deployable fully disconnected, with all AI processing local to your infrastructure.
Open: an open API and APPs ecosystem let your agency and its trusted integrators add capability independently of the vendor.
All-source by design: one ontology and one graph across SIGINT, OSINT, HUMINT, FININT and cyber data.
Petabyte scale: a ClickHouse-based engine delivers sub-second search over structured and unstructured collection.
Governed AI: transparent outputs, source traceability and audit logging, aligned with EU and international AI frameworks.
Collaboration with BAE Systems announced in 2026; in use with national police.

FAQ

Frequently asked questions

Can Octostar be accredited for classified networks?: Octostar is designed to run inside an agency's own accreditation boundary. It deploys on-premise or air-gapped, has no dependency on external services, performs AI processing locally and provides the audit logging and access controls that accreditation processes require. Accreditation itself is carried out by the agency under its national scheme.
Which intelligence disciplines does it support?: The platform is discipline-agnostic. It has been applied to SIGINT, OSINT, HUMINT, FININT, ADINT and cyber data. Each source is mapped onto the shared ontology, so analysis works the same way regardless of where the data came from.
Do the AI models require a connection to an external provider?: No. Transcription, translation, image and video understanding, entity extraction and the generative assistant all run on local models inside your deployment. The Octobox appliance demonstrates this in a fully disconnected form.
How does Octostar compare with Palantir Gotham?: Both are all-source analysis platforms built around a knowledge graph. Octostar differs in being European-built, deployable and extendable independently of the vendor through an open API and APPs, and available on an appliance for air-gapped use. Agencies evaluating alternatives to US platforms typically start with an Octobox proof of concept on their own data.
Can partner agencies share a deployment?: Yes. Compartmented access control, per-record permissions and shared workspaces allow several units or partner agencies to work on one deployment while each sees only what it is cleared to see.