Sovereign AI for Canberra and Federal Government

Yes. PROTECTED-accredited deployment is achievable through a combination of deployment architecture aligned to the relevant ISM controls, an IRAP assessment by an ASD-endorsed assessor, and integration into your agency's accreditation process. The deployment must run on PROTECTED-certified infrastructure, typically within the agency's own data centre or on a cloud platform with a current PROTECTED certification from ASD. We provide the deployment documentation, ISM control mapping, and technical specifications required to support an IRAP assessment.

For SECRET and above environments, the appropriate architecture is a fully air-gapped deployment with no external network connectivity whatsoever. This means the LLM runs on hardware physically located inside your secure compartmented information facility, with all data staying within that environment. Model updates are delivered through your existing secure supply chain rather than over a network connection. We have deployed in environments at this classification level and can work with your cleared facilities team and security authority to design and implement an appropriate solution.

Custom LLM deployments for Australian government agencies can be procured through several pathways. For initial engagements, the Digital Transformation Agency's Digital Marketplace and the Commonwealth Government's ICT Procurement frameworks provide relevant procurement vehicles. For larger deployments, a Request for Tender process is appropriate. We are experienced with the whole-of-government procurement frameworks and can work within your agency's delegation thresholds. For agencies preferring a staged approach, an initial pilot under a lower-value procurement can demonstrate value before a larger rollout.

APS mobility is one of the most significant knowledge management challenges in the public sector. When an SES officer departs or transfers, they take with them the institutional memory of program decisions, stakeholder relationships, ministerial preferences, and the reasoning behind current policy settings. A custom LLM trained on your agency's policy documentation, briefing history, and program records creates a searchable institutional memory that survives individual transitions. Incoming SES officers can query the model to understand the history and rationale of programs they have inherited, dramatically reducing the ramp-up period.

For ACT government agencies and organisations operating under ACT legislation, the model can be trained on ACT Government-specific frameworks including the Human Rights Act 2004, the Information Privacy Act 2014, the Territory Records Act, the Crimes Act, and sector-specific ACT legislation for health, education, and housing. For Commonwealth agencies in Canberra, the model covers the full suite of Commonwealth legislation and administrative frameworks including the PGPA Act, the Archives Act, the FOI Act, and the Privacy Act. Both frameworks are maintained with regular updates.

As of 2025, there is no single whole-of-government procurement vehicle specifically for sovereign AI deployments, though the Digital Marketplace and various agency-specific ICT panels are available. We recommend agencies consider a two-stage approach: an initial proof of concept under a standing panel arrangement, followed by a larger deployment contract procured through open tender or a limited tender process from demonstrated suppliers. We can assist with the specification development for the second stage based on learnings from the proof of concept.