Artificial intelligence, explained.
Australia's Prime Minister Anthony Albanese outlines his government's approach to AI on 15 July 2026 (David Gray/AFP via Getty Images)
Albanes’s speech this week marks a real advance. But Australia needs governance to match its ambition.
Subscribe to The Informer for monthly expert analysis, and to Events for advance notice of visiting world leaders and distinguished guests.
You may unsubscribe from Lowy Institute newsletters at any time. For information on our privacy practices and how to unsubscribe, see our Privacy Policy.
The Prime Minister has opened the right AI debate – what Australia must do nextThe Prime Minister has opened the right AI debate – what Australia must do next
Prime Minister Anthony Albanese has shifted Australia’s artificial intelligence debate onto the ground where it always belonged. AI is no longer being presented simply as another technology to adopt, another productivity tool to promote or another industry seeking government support. In his keynote address (Opens in new window) at the University of Sydney, Albanese placed it within the larger questions of sovereignty, national resilience, economic security and Australian values.
That matters because Australia’s AI debate has too often been trapped between two shallow positions. One side promises investment, productivity and economic growth; the other warns of job losses, bias and technological harm. Both contain truth, but neither amounts to a national strategy.
The more important question is whether Australia will shape the systems upon which it increasingly depends or merely accept systems designed elsewhere, according to rules written by others.
Last year, I argued (Opens in new window) in The Interpreter that Australia faced a choice between becoming an AI standards setter and remaining a technology taker. If the rules governing AI were written offshore, Australia would import not only the technology but also the assumptions, interests and priorities built into it. I later argued (Opens in new window) that the government’s National AI Plan (Opens in new window), with its reliance on existing technology-neutral laws, risked replacing strategy with caution. Existing legislation could deal with some harms, but it was not designed for systems that operate at scale, obscure responsibility and influence decisions across employment, security, government and public life.
The Prime Minister’s speech does not resolve those concerns, but it shows that the government now understands them more clearly.
Albanese warned that Australia must become more than a “data warehouse for AI products made overseas”. He argued that the country needs a stake in where AI is made, how it is made and the rules under which it operates. He also acknowledged that dependence on foreign technology can become a source of national vulnerability. The government is no longer speaking only about adopting AI – it is speaking about shaping it. The announcement of Australian AI standards, a national framework for large data centres and a new Office of AI within the Department of the Prime Minister and Cabinet shows that the issue has moved closer to the centre of government. The direction is right. The next task is to define what these commitments mean in practice.
Values without technical competence become slogans, while technical competence without a governing purpose becomes capability in search of permission.
A data centre on Australian soil is not necessarily sovereign capability. The models, the intellectual property and the most important technical knowledge may remain overseas, and decisions about access, pricing, security and future development may still be made in foreign boardrooms. Australia could end up supplying the land, electricity and water while others retain the technology, knowledge and strategic control.
Foreign investment is necessary and should be welcomed, but investment and approval settings should also be used to build lasting Australian capability. It should support Australian research, workforce development, domestic access to computing power and participation by Australian companies. Government must also identify the capabilities that are too important to leave entirely in foreign hands. Sovereignty does not require Australia to build everything itself. It does require us to know what we must control, what we can safely share and what dependencies we are prepared to accept.
Albanese was right to defend Australian writers, artists, musicians and journalists, and his statement that creative work should not be taken without the creator’s control, including control over price and value, draws a clear moral line. But the standards must also deal with the systems themselves, particularly those used in government decisions, workplaces, health, justice, Defence, critical infrastructure and other areas where errors can cause serious harm.
The question is not whether these systems should be banned, but what evidence should be required before they are trusted. It is not enough to place a person somewhere in the process and describe the arrangement as human oversight. That person must understand the system well enough to question it, possess the authority to reject its recommendation and remain accountable for the final decision.
Albanese said that the central issues raised by AI are economic, legal, social, moral and spiritual rather than simply technical. He is right that AI cannot be left to engineers and technology companies alone, yet those wider issues cannot be separated from technical design. Bias, opacity, manipulation and loss of control do not appear in the abstract; they emerge from the way systems are trained, tested, deployed and connected to human decisions. Australia needs people who can work across the boundaries between technology, law, policy, ethics and operations. It must avoid the equally dangerous mistakes of allowing technical specialists to define the public interest or asking generalist policymakers to govern systems they do not understand. Values without technical competence become slogans, while technical competence without a governing purpose becomes capability in search of permission.
Professional judgement should therefore be treated as a national capability. AI should strengthen it, not replace it. This does not mean preserving every old job or process. But productivity cannot be measured only by how much labour or time can be removed from a process. It must also account for what knowledge, responsibility and resilience may be removed with it. An organisation that produces decisions more quickly while losing the capacity to understand them has not necessarily become more capable.
About the author
Ian Gribble
Ian Gribble is an independent strategic advisor and writer based in Canberra.
The most-pressing world events explained by Lowy Institute experts and global contributors, in your inbox, every Thursday.
You may unsubscribe from The Interpreter at any time. For information on our privacy practices and how to unsubscribe, see our Privacy Policy.
