For the better part of two years, the criticism of Canberra’s approach to artificial intelligence has been remarkably consistent: plenty of consultation, plenty of principles, and very little in the way of a decision. Anthony Albanese‘s first major speech dedicated to the technology was meant to change that, and in one important respect it did. It put AI where informed observers have long argued it belongs, at the centre of the national economic and security conversation. Yet a fresh analysis from the Australian Strategic Policy Institute argues the speech left the hardest question almost entirely unanswered. What, exactly, does Australia want to be able to do for itself, and how does it plan to get there?
The news
Writing for the ASPI Strategist, the institute welcomed the elevation of AI to a genuine national priority while warning that the Prime Minister’s plan for sovereign capability remains, in its assessment, unclear. The through line of the ASPI piece is that recognising AI as a priority is necessary but nowhere near sufficient. A speech can declare that the technology matters. It cannot, on its own, tell industry which parts of the AI stack the country intends to own, which it is content to import, and where public money and political capital will actually be spent.
The distinction matters because sovereign capability is one of those phrases that can mean almost anything. For some it means domestic data centres and the compute that runs inside them. For others it means Australian-trained foundation models, or local control over the data used to train them, or simply the skills and institutions needed to deploy overseas models safely on Australian terms. ASPI’s point, as Anthony Albanese steps up the government’s rhetoric, is that the country cannot build toward a goal it has not defined.
Two ways of reading the speech
There are, broadly, two ways to interpret what the Prime Minister delivered. The generous reading is that this is a deliberate sequencing. Governments rarely announce the architecture and the budget in the same breath as the ambition, and the recently created Office of AI gives Canberra an institutional home from which a more detailed strategy can flow. On this view, the speech was a signal to markets, universities and the public service that AI is now a first-order priority, with the substance to be filled in as the office finds its feet.
The less generous reading, and the one ASPI leans toward, is that the absence of specifics is not a matter of timing but of unresolved internal disagreement. Australia’s AI debate has been pulled in competing directions for months. There is the copyright fight over whether local creative and media content can be scraped to train models. There is the surge of data centre proposals straining the electricity grid. There is the tension between wanting to attract global hyperscalers and wanting something recognisably Australian to show for it. A speech that avoids picking sides on any of these can look statesmanlike, or it can look like a government that has not yet made up its mind.
Neither reading is entirely fair on its own, which is rather the point. The government would argue it is moving deliberately in a fast-moving field where locking in the wrong bet is expensive. Critics such as ASPI would counter that deliberation has already consumed the better part of a term, and that other countries have stopped consulting and started building.
Why this lands so hard in Australia
For a mid-sized economy, the sovereign AI question is not academic. Australia does not have the domestic market or the capital base to match the United States or China across the whole technology stack, so every dollar and every policy lever has to be pointed at something specific. Get the choices right and the country builds durable advantage in areas where it already has strengths, from resources and agriculture to health data and defence. Get them wrong, or make none at all, and Australia ends up as a well-serviced customer of other people’s systems, paying rent on infrastructure and models it neither controls nor fully understands.
That is the anxiety running beneath the sovereignty debate, and it is why the vagueness ASPI identifies carries real cost. Industry cannot invest against a strategy that has not been written down. A start-up deciding whether to train a model in Australia, a super fund weighing a data centre, a university choosing where to place its research bets, all of them need to know what the government will back and what it will leave to the market. Ambition without specificity leaves each of them guessing, and guessing tends to favour the safe option of buying from overseas.
There is also a security dimension that ASPI, as a strategic policy institute, is well placed to press. Sovereign capability is not only about economic upside. It is about whether Australia can assure the integrity of the AI systems embedded in critical infrastructure, government services and the defence enterprise. On that measure, hosting someone else’s technology is not the same as being able to build, audit and, if necessary, replace it.
What happens next
The immediate test is whether the Office of AI turns the Prime Minister’s framing into decisions. That means answering the questions the speech sidestepped: a working definition of sovereign capability, a sense of which layers of the stack Australia will prioritise, and some indication of the money and the timelines attached. Until those arrive, the gap ASPI describes will keep drawing critics from across the spectrum, including those who think the government is doing too little and those who think it should get out of the way and let private capital sort it out.
Albanese has at least closed one debate. AI is now unarguably a national priority, spoken about from the top in the language of economic security. The harder debate, about what Australia actually intends to build with that priority, is only beginning. On the evidence of this speech, the answer is still being worked out somewhere inside government, and the country is being asked to take it on faith that a plan exists. ASPI’s polite but pointed conclusion is that faith is not a strategy.
Sources: ASPI Strategist.
