Defence and national security agencies
Teams that need AI on sensitive data, inside the environments and under the security policies they already run.
Defence AI that stays under your control
We design, build and implement defence AI for agencies, national security teams and the defence industry: secure, sovereign systems that run inside your environment, on-premise or fully offline, and support the people who make the decisions.
- Runs where you need itOn-premise, in your tenancy or air-gapped
- People decideAI that supports analysts and commanders
- You own the workCode, models, data and documentation
- NDA before discoveryMutual, and we are happy to sign yours
What is sovereign defence AI?
Sovereign defence AI is artificial intelligence that a defence or national security organisation controls end to end: the data, the models, the infrastructure they run on and the people who can reach them. In practice that means open-weight models and search indexes deployed inside your own environment, often with no connection to the internet, so nothing sensitive leaves your control.
We build these systems for decision support, sustainment, analysis and security, and we move the legacy platforms underneath them to supported technology. Every engagement has a fixed fee agreed in writing, and you own what we create.
What we build
Capabilities described in general terms. Client work is confidential.
Secure knowledge assistants
Cited answers from manuals, procedures, reports and lessons learned, running offline on your own hardware.
Private AI assistants
Sustainment and logistics
Predictive maintenance and spares forecasting for vehicles, vessels, aircraft and equipment, measured against today's availability.
Decision intelligence
Imagery and sensor analysis
Anomaly and change detection that points human analysts to what needs a closer look.
Custom algorithms
Planning and decision support
Optimisation for schedules, rosters, routes and resource allocation, with people approving every recommendation.
Custom algorithms
AI assurance and red teaming
Models, assistants and agents tested for data leaks and manipulation before they connect to anything sensitive.
AI security testing
Quantum-safe cryptography
An inventory of vulnerable cryptography and a staged plan to meet ASD's guidance to stop using traditional public-key algorithms by the end of 2030.
Quantum-safe readinessWho we work with
Defence industry
Primes, small and medium enterprises and suppliers building capability, sustainment and logistics systems for defence customers.
Research and dual-use programmes
Universities and start-ups turning research in AI, quantum computing and robotics into working prototypes.
Legacy system migration
Defence platforms that have outlived their support
Old databases, mainframe and IBM i applications and unsupported servers moved to supported platforms in fixed-price 30-day sprints, inside your environment, with a parallel run and a rehearsed rollback.
from A$50,000 per sprint
Built for secure environments
Inside your environment
We build and run in your network, your tenancy or an air-gapped enclave, under your security policies.
Offline when required
Open-weight models, local embeddings and a local search index, with no calls to outside services.
People make the decisions
Systems recommend and explain. People review, approve, override or switch them off.
Logged and documented
Every question, source and output logged in your environment, with controls designed with the ASD Essential Eight in mind.
Tested before connected
Adversarial testing for prompt injection, data leaks and excessive agency before anything goes live.
Access for the engagement only
Least-privilege named access, removed and confirmed in writing when we finish.
Ways to start
Each engagement has a defined scope and a fixed fee agreed in writing before it starts.
Prices in Australian dollars, excluding GST. Scope and fee are confirmed in writing after a free call.
Do you work with defence organisations?
Yes. Defence is one of the sectors we have worked in. We describe that work only in general terms and do not publish details of client engagements.
Can the AI run without an internet connection?
Yes. For sensitive environments we deploy open-weight models, embeddings and the search index entirely on your own hardware, including fully offline and air-gapped setups. Our lab work includes an offline knowledge assistant that runs on consumer hardware with no cloud dependency.
Do you hold security clearances or DISP membership?
If your work needs a particular clearance, Defence Industry Security Program (DISP) membership or another accreditation, raise it on the first call and we will tell you plainly whether we meet it.
Will the AI make decisions on its own?
No. We build decision support. Systems recommend, rank and flag, explain why, and leave the decision with the people accountable for it, with clear override and shutdown paths.
How can we start without sharing sensitive data?
With a confidential call and a mutual NDA. Scoping, architecture and evaluation design can use synthetic or non-sensitive samples. Work on real data happens inside your environment, under your controls.
Who owns the models and code?
You do. Deliverables and source code created for you are assigned to you on payment. We keep our pre-existing tools and methods and grant you a perpetual licence to any that are embedded in your solution.
Start with a confidential call
Tell us the capability you need and the environment it must run in. We reply within one business day, and we are happy to sign an NDA first.
Every engagement is led by our principal and staffed with the relevant professionals your goal needs.
Prefer email?
contact@sovereignsystemslabs.com