SOVEREIGN & HIGH-ASSURANCE

Run it inside your perimeter. With AI, or without.

Camlin Connect runs on shared Sydney cloud today, and is delivered with your team into your own cloud or an air-gapped network. A validated No-AI mode refuses AI steps when you save and when the service runs. Identity plugs into the OIDC provider you already run. The speech stack is Australian-built with documented lineage. Built by an Australian vendor with years of production operations inside a high-security network environment.

Request the Sovereign Assurance PackTalk to the deployment team
The whole conversation stays inside your perimeter, and the AI has an off switchA dashed boundary labelled Your perimeter, with a padlock on it. Inside: a caller on the phone, the avatar on a mobile taking his call, and a switch labelled AI off, AI steps refused. The conversation runs inside the line.YOUR PERIMETERAI offAI steps refusedRuns inside your network

Deploy without AI today. Enable it per step type as Australian frameworks land.

Australia’s AI governance is moving quickly — the Office of AI stood up in July 2026, with mandatory AI standards expected to follow. Camlin Connect decouples the decision: launch deterministic voice and chat services now, then switch on AI capabilities step type by step type, on your timetable, as your agency’s policy settles. No re-platforming, no rebuild.

NO AI IN RUNTIME

AI steps refused. Enforced, not promised.

Not a policy promise: an enforced platform mode, validated end to end in July 2026. A service flagged No AI in runtime cannot save or run an AI-bound step, and the check happens twice.

A closed list of AI-bound steps

Every step type in a service Blueprint is classified. Three step types are AI-bound (AI intent, AI sentiment, AI generate) and refused in No-AI mode. Menus, forms, intent matching from phrases, data capture, routing and integrations keep working.

Rejected at save time

Mark a Blueprint "No AI in runtime" and the platform refuses to save any version containing an AI-bound step. Non-compliant designs cannot reach production in the first place.

Enforced twice

Even if a non-compliant flow somehow reached the runtime, an AI-bound step is refused before it runs. No-AI mode was validated end to end with no AI provider configured.

Design-time AI stays separable: teams can use AI to draft and review Blueprints while the running service keeps AI-bound steps switched off. How No-AI mode works →

DEPLOYMENT KINDS

From shared Sydney cloud to air-gapped

Every tenant declares one of four deployment kinds, recorded on the tenant itself, so the deployment posture cannot drift at runtime. Shared and dedicated are live; your cloud and offline installs are delivered with your team.

Shared cloud

Shipped

Multi-tenant SaaS in AWS ap-southeast-2 (Sydney). The fastest path to a working service. Data stays in-region.

Dedicated

Shipped

Single-tenant infrastructure, isolated compute and storage, operated by Nuamedia. For teams that need separation without running it themselves.

Customer cloud

Delivered with your team

The platform installed inside your own cloud account, under your controls, your network boundaries and your monitoring. Planned and delivered with your team.

Offline / air-gapped

Delivered with your team

For networks that cannot depend on the internet. Licences are Ed25519-signed files verified locally. No activation call, no telemetry required to operate. Planned and delivered with your team.

IDENTITY & CONTROL

Your identity provider. Your keys. Your controls.

High-assurance environments already run their own identity, their own keys and their own change control. The platform meets them there instead of asking them to move.

Standards-based OIDC

Shipped

Token validation is standard OIDC/JWT via JWKS. Keycloak, Okta, Auth0, the major cloud identity services and any other compliant issuer plug in — including identity providers running inside an air-gapped network.

Keycloak claim preset

Configurable

Pluggable claim mappers ship with Keycloak and standard-OIDC presets, so self-hosted identity maps onto platform roles without custom code.

Roles stay in the platform

Shipped

Owner, admin, editor and viewer roles are managed in-platform, independent of the identity provider. Swapping IdPs never rewires who can do what.

Bring your own AI keys

Shipped

Where AI is enabled, AI steps run on your own tenant-scoped keys, so the commercial and data relationship with the AI and speech providers stays yours.

The voice layer is sovereign too

Camlin Speech is an Australian-built speech stack — Australian-accent recognition, neural text-to-speech and avatar-ready viseme output — with documented model lineage and no Chinese-origin weights in the production tree. It runs in Sydney today; running it on your own hardware, with the audio path inside your perimeter, is delivered with your team.

How Camlin Speech maps against PSPF, FOCI and HCF →
PROVEN WHERE IT COUNTS

Heritage in high-security networks

This isn't a compliance page written ahead of the product. The delivery pattern behind it has run for years in environments that don't tolerate failure.

High-security network operations

Years of production operations inside a high-security network environment. The operational discipline behind the platform’s air-gapped deployment model. Details are available under NDA as part of the assurance pack conversation.

Public-sector delivery

Wheelchair Book & Ride, Sydney’s booking service for wheelchair-accessible taxis, operated by SCIA and funded by Transport for NSW, runs on this platform — thousands of monthly bookings, WCAG 2.1 AA, delivered on time. Read the case study →

REQUEST THE PACK

Get the Sovereign Assurance Pack

The Sovereign & High-Assurance Deployment Pack covers the four deployment kinds, No-AI runtime evidence, identity and licensing detail, and the speech supply-chain lineage, with every claim labelled by status and mapped back to the codebase.

We’ll respond within one business day with the pack and offer a 30-minute walkthrough with the deployment team.