Sovereign AI gateway
Sovereign AI, live in Australia today.
A secure, fully audited AI gateway for classified and regulated workloads. Built, operational, and ready to scale, with every byte resident onshore.
The problem
Compliance is not optional. Sovereign AI is hard to operationalise.
Government, defence, financial services, and health face mandates that dictate where data goes and who can touch it.
ISM, PSPF, and sector-specific frameworks are not guidance, they are obligations. The bar is to demonstrate compliance, not merely claim it.
Sovereign AI exists on the hyperscalers. Operationalising it for accreditation is where most organisations stall: building the security boundaries, audit controls, and operational processes that survive an IRAP assessment.
Model availability varies by region and provider. Regulated organisations need the best models available, not just what one cloud happens to offer.
The platform
One sovereign gateway. Chat, CLI, and API.
Built for organisations that must prove their controls. It slots into existing infrastructure as a discrete, assessable block, without replacing the tools and workflows you already run.
Accreditation-ready
Well-defined security boundaries that minimise the friction of an IRAP assessment. ISM and PSPF, to PROTECTED.
Fully sovereign
Runs entirely in Australian cloud regions. Your data never leaves the country.
Fully audited
Every interaction logged, encrypted, and attributable. Customer-managed encryption keys throughout.
Multi-model by design
The leading frontier models, served through one OpenAI-compatible gateway. Model-agnostic and cloud-agnostic by architecture, never locked to a single provider.
Native Claude Code
The leading agentic coding assistant, running natively through sovereign infrastructure. Fully compliant.
Zero friction
A developer CLI and an OpenAI-compatible API. Drop-in alongside the workflows your teams already use.
Modular
A discrete, assessable component that integrates with existing systems rather than displacing them.
Secure by design
Your data stays in Australia. By design, not by promise.
Sovereignty and data residency are the reason Kelios exists. The controls are built into the architecture, and KPMG retains full visibility over how AI is used on your work.
Onshore, always
Inference, storage, and logs run in Australian cloud regions (ap-southeast-2). No data path leaves the country.
Customer-managed keys
Encryption keys you hold and control. Data at rest and in transit is encrypted end to end.
Attributable and audited
Every request is logged and tied to an identity. Who used what, when, and why, for evidence and review.
Locked-down boundary
A contained, in-region perimeter with no general internet egress on the model path. Governed access, gated by identity.
Global scale
One architecture, any jurisdiction.
Sovereign AI is not only an Australian need. Kelios is region-agnostic: it deploys into any jurisdiction's sovereign cloud as a fully isolated, accreditation-ready capability. The same architecture, wherever your obligations live.
Delivery models
Hosted by us, or deployed into your cloud.
01 / Internal
KPMG Internal Secure AI
Secure AI for classified engagements and sensitive client work, run by KPMG.
02 / Managed
Managed Service
KPMG hosts, your teams use. No client infrastructure to stand up or maintain.
03 / Packaged
Packaged Deployment
Deployed into your own cloud as a consulting engagement, licensed for the long term.
04 / Air-gapped
Secure-AI-in-a-Box
Dedicated, air-gapped hardware for the highest classification levels.
Built and operational, today
Sovereign AI you can put in front of an assessor.
Talk to the team about a briefing, a managed instance, or a deployment into your environment.