Unclassified // Public release
Security architecture of SynapseWargame™
SynapseWargame™ is built for environments where a security compromise is an operational failure. Every layer - data handling, cryptography, deployment topology and the AI stack itself - is engineered for sovereign, air-gapped defence use at up to NATO SECRET.
Classification-aware by design
Classification is handled per data element, not per system. Every order-of-battle entry, assumption, simulation result and recommendation carries its own classification and releasability marking, and those markings propagate through derived products automatically - a comparison matrix inherits the highest marking of the evidence that produced it.
Releasability caveats are respected throughout: NATO SECRET (NS), NATO CONFIDENTIAL (NC) and national eyes-only material remain visible only to cleared, entitled users. In coalition planning, partitioned views mean thirty-two nations can work the same scenario while each sees only what it is entitled to see, with Alliance-layer products assembled from the releasable subset.
Post-quantum cryptography on every layer
Decent Cybersecurity is a post-quantum security firm first. SynapseWargame™ applies PQC across all three layers: transport between components and enclaves, storage of scenarios and evidence at rest, and the signatures that make the evidence chain tamper-evident.
Signing is not an export-time afterthought. Inputs, assumptions, simulation runs and recommendations are signed as they are created, so the chain reflects the decision as it actually unfolded. Records signed today remain verifiable against a future adversary with quantum capability - a requirement for material that will be scrutinised for decades.
Deployment models
The same capability is delivered in three topologies. Air-gapped is not a reduced tier - the disconnected enclave runs the complete platform.
| On-premises | Sovereign cloud | Air-gapped enclave |
|---|---|---|
| Hosted in customer data centres under customer physical control | Cleared European infrastructure, no foreign operator access | Zero external connectivity; no route to the public internet at any time |
| Full capability, including Red Agent and signed evidence chain | Full capability with elastic simulation capacity for exercise peaks | Full capability - nothing is withheld from the disconnected configuration |
| Classification ceiling set by the accredited facility, up to NATO SECRET | Suitable up to EU SECRET subject to accreditation of the hosting enclave | Designed for NATO SECRET and national eyes-only working environments |
| Updates delivered as signed packages through customer change control | Continuous delivery of model and scenario library updates | Offline signed media transfer for updates and scenario libraries |
Sovereign AI stack
Models are fine-tuned for staff planning language and operated on cleared European infrastructure. There is zero dependency on foreign API providers: no prompt, scenario or fragment of order-of-battle data leaves the deployment boundary, and no capability degrades when connectivity is removed.
Model behaviour is versioned alongside the evidence chain. A wargame run six months ago can be replayed against the exact model version that produced it - sovereign military AI in Europe means controlling not only where inference happens, but which artefact performed it.
Assurance
Decent Cybersecurity holds ISO 27001 and ISO 9001 certification, with facility and personnel clearances at NATO SECRET and EU SECRET. It is a registered NATO supplier (NSPA/NCIA) and was selected for the NATO DIANA Decision Superiority cohort.
Assurance extends into the product. Session replay reconstructs any wargame exactly as it happened, and the signed evidence chain lets an inspector-general, legal advisor or auditor verify independently that the record has not been altered since the decision was taken.
Request the capability brief for the full security annex, or request a classified briefing.