Home / Security

Your estate data, and your reasoning, stay yours.

Critical national infrastructure can't run on public-cloud chatbots that leak data and hallucinate. Port-Systems is built around four hard constraints those tools fundamentally can't meet, enforced by the neuro-symbolic architecture itself.

Public-cloud chatbots vs Port-Systems

Same surface. Very different consequences.

Public-cloud chatbot

Generative, multi-tenant

Port-Systems

Neuro-symbolic, sovereign

Data residency
Often non-UK, multi-tenant
UK / private cloud, single-tenant
Data handling
Prompts and data may train external models
Your data never leaves your boundary
Auditability
Black-box generation, no derivation
Symbolic derivation; every answer cites its source
Hallucination risk
Fluent but unverifiable answers
Answers derived by formal rules over a graph
Access control
One account, one view
Role-based, Finance ≠ Operations
Domain logic
Implicit in model weights, drift-prone
Explicit ontology and rules, versioned, reviewable

The four pillars

How "sovereign" is enforced, not just claimed

Data residency

Hosted in UK / Scottish private cloud. Your operational records, your graph and your reasoning don't leave your jurisdiction, by architecture, not just policy.

No external sharing

Queries and data are never shared with, retained by, or used to train external services. The neural interface runs against your governed graph, not the open web.

Symbolic auditability

Every answer comes with its derivation: the rules applied, the entities traversed and the source records used. Lineage is visible to the user, not just the auditor.

Role-based access

Finance sees finance. Operations sees operations. The same platform serves the boardroom and the dock without leaking either way.

Risks we plan for

No deployment is risk-free. We name them up front.

Data fragmentation

Staged integration into the knowledge graph, exports first, live APIs second. We don't block on perfect connectivity.

LLM hallucination

The neural layer compiles intent into formal queries, it never authors the answer. Outputs come from the symbolic engine, with mandatory citations.

User adoption

Champion user groups in each department, with the derivation chain shown alongside every answer for trust by construction.