AI sovereignty is rapidly becoming a strategic risk issue, not a policy ambition. This discussion with Dr. Jon Whittle – Former Managing Director of Data61 at CSIRO and Simon Kriss – CEO at Sovereign AI Australia, reframes sovereignty as a multi‑layered operating concern that affects national resilience, organisational accountability, and long‑term economic value, well beyond where data happens to be hosted.

Sovereignty extends across the entire AI stack

True AI sovereignty cannot be reduced to local data centres or Australian cloud regions. It spans six interdependent layers: data, foundation models, inference, agentic systems, responsible AI, and applications. Weakness at any layer undermines the whole. Treating sovereignty as a partial solution, focused only on storage or applications, creates a false sense of control while leaving critical dependencies embedded elsewhere.

Foreign dependence introduces compounding risk

Australia increasingly relies on offshore foundation models, inference infrastructure and agentic tooling, exposing sensitive data, national values and economic flows to external control. Legal realities such as foreign government access laws, value misalignment, and unaccountable training data challenge explainability, compliance and trust. As AI systems move closer to decision‑making and autonomy, these risks become operational rather than theoretical.

The next AI race is still open

While Australia missed the starting line for large language models, the next phase of AI; neurosymbolic systems, world models and energy‑efficient agents, remains undecided. Sovereignty therefore becomes a forward‑looking capability play investing early in skills, models, agent platforms and data governance aligned to Australian law and values. Organisations that act now improve resilience, optionality and credibility before the issue is forced by incident, regulation or public scrutiny.

Key takeaways:

• Sovereignty is systemic: Control must exist across data, models, inference, agents and governance—not just infrastructure location.
• Risk compounds silently: Offshore AI dependencies accumulate legal, ethical and economic exposure long before failure is visible.
• The next race matters more: Australia can still lead by investing in the post‑LLM era rather than replaying the last one.