Australia’s Defence department says its AI adoption has accelerated over the past three months, as its chief AI officer works to bring the department’s efforts into line. The account offers a concrete look at how a large government organisation is building an AI portfolio and sourcing access to frontier models.
Watch Desk analysis
What happened
Defence chief AI and information officer Chris Crozier told a parliamentary committee that he and colleagues had interviewed more than 60 staff actively applying AI. He said the work had built a comprehensive portfolio of Defence’s engagements, spanning military and business-productivity uses.
Crozier said Defence currently relies on the Australian Signals Directorate to broker access to frontier models. He argued those models matter for national sovereignty, while adding that a good open-source model paired with effective software scaffolding can also deliver a strong result. The account comes from iTnews’ report on the committee hearing.
Why it matters
This is more than a general claim that government wants to use AI. Defence is trying to understand where the technology is already being applied, coordinate that work and scale useful results. Its reliance on the Signals Directorate also shows how access to frontier systems can be shaped by national security and sovereignty concerns.
Crozier’s comments describe the department’s approach, not independently measured outcomes. Still, the combination of staff interviews, central coordination and a route to both frontier and open-weight models makes the practical questions unusually clear: what gets deployed, and how does it serve Defence’s operational needs?
Our read
The most useful signal here is the attempt to map existing work before scaling it. Sixty interviews will not, by themselves, prove that AI is improving Defence, but a coherent view of who is using it and for what is a better starting point than sprinkling models across a department and hoping for the best.
Keeping open-source models in the picture is also notable. Frontier access may be important, as Crozier argues, but the department is not treating one class of model as the only route to useful results. The test now is whether that portfolio turns into specific, measurable deployments.
What to watch
- Whether Defence publishes examples of AI applications and the results they deliver.
- How the department evaluates frontier and open-weight models for different tasks.
- What role the Australian Signals Directorate plays in providing or brokering model access.
Discussion spark: Should a defence department prioritise access to the most capable frontier models, or build more of its AI work around open-weight systems it can control?
Sources and evidence
- Source update (1 November 2026, 00:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.