For a policy area that touches everything from the tax office to the classroom, Australia’s approach to artificial intelligence has stayed remarkably unsettled. That tension was captured neatly in a recent segment from the national broadcaster, which asked a question plenty of executives, public servants and union officials have been muttering for months: is Australia’s AI policy about to change?
It is a fair thing to ask. The Commonwealth has spent the past two years signalling intent without landing on a fixed regime. It released a voluntary AI Safety Standard, floated the idea of mandatory guardrails for high-risk uses, stood up a National AI Centre, and leaned heavily on the Productivity Commission to work out how much economic upside is actually on the table. What it has not done is legislate a comprehensive AI act in the mould of the European Union. For businesses trying to plan multi-year investments, and for citizens wondering who is accountable when an automated decision goes wrong, that ambiguity has real costs.
How Australia got here
The current settings trace back to a 2023 consultation on safe and responsible AI, which drew hundreds of submissions and made one thing clear: existing laws covering privacy, consumer protection, discrimination and workplace safety were never designed with generative systems in mind. The government’s response leaned towards a risk-based model, promising tougher obligations for high-stakes applications such as those in health, employment and law enforcement, while leaving lower-risk uses to run under existing rules.
Since then the pace has been deliberate rather than dramatic. A proposals paper canvassed making guardrails mandatory in high-risk settings, including testing, transparency and human oversight requirements. The National AI Centre, now housed within the industry portfolio, has pushed voluntary standards and adoption support. And the change of ministerial faces after the 2025 election added another variable, with responsibility for the agenda sitting with Tim Ayres as the minister overseeing industry, innovation and science. The direction of travel has been consistent, but the destination keeps moving.
Two camps, pulling hard
The reason the question feels live is that the pressure is coming from opposite directions at once.
On one side sit the technology vendors, the big consultancies and much of corporate Australia, who argue that heavy-handed rules would ka-put the productivity gains before they arrive. Their case is straightforward: AI is one of the few genuine levers left to lift a stagnant productivity rate, and regulatory uncertainty is itself a handbrake because boards will not sign off on serious spending when the rules could shift under them. This camp tends to favour light-touch, principles-based settings, harmonised with major trading partners so Australian firms are not forced to build bespoke compliance for a market of 27 million people. They point to the risk of exporting jobs and data centres offshore if the local regime becomes an outlier.
On the other side are unions, consumer advocates, human rights bodies and a good slice of the research community, who worry that voluntary standards are no substitute for enforceable rights. Their concern is not abstract. Automated systems already sit inside hiring, credit, insurance, welfare and policing decisions, and the memory of the unlawful Robodebt scheme still shapes how many Australians view government automation. This camp wants mandatory guardrails, clear liability when things go wrong, and meaningful transparency so people can contest a decision made partly or wholly by a machine. To them, a change in policy that softens or delays those protections would be a step backwards, not forwards.
The government is trying to hold both truths at once, which is precisely why the policy has felt stuck. It wants the investment and the jobs, and it does not want to be the administration that presided over the next automated-decision scandal.
Why the stakes are distinctly Australian
It would be easy to treat all this as a local echo of a global debate, but the Australian context has its own sharp edges. The economy is heavily weighted towards services, resources and public administration, all of which are exposed to automation in different ways. The country is also a net importer of the underlying models, which means most of the frontier technology, and most of the compute, is controlled offshore. That raises the sovereignty questions that keep surfacing across the sector, from defence planners worried about subsea cables and autonomous systems to procurement officials debating whether government should build local large language models rather than rent foreign ones.
There is a federation problem too. States are moving at their own speed on issues such as facial recognition, education and the use of AI in the public service, which risks a patchwork that is harder for national employers to navigate than a single Commonwealth framework would be. And the workforce dimension is unavoidable. Australians have watched a string of companies attribute job cuts to a mix of cost pressure and automation, and that lived experience colours how any policy change will be received. A regime seen as friendly to employers but silent on workers will struggle for legitimacy, while one that spooks investment will be blamed for every project that heads overseas.
What could actually shift
So is change genuinely coming? The most likely near-term move is not a single grand AI act but a set of targeted interventions layered onto existing law. That could mean turning parts of the voluntary safety standard into mandatory obligations for defined high-risk uses, tightening privacy reform so it covers automated decision-making more explicitly, and giving existing regulators sharper tools rather than creating an entirely new AI agency. The Productivity Commission’s work will be central here, because it gives the government the economic cover to argue that sensible rules and strong growth are not in conflict.
The timing is also political. With the productivity conversation dominating the economic agenda and a steady drumbeat of AI incidents in the courts and the headlines, the government has both the incentive and the pretext to act. The risk it is weighing is that moving too slowly cedes the field to a patchwork of state rules and public anxiety, while moving too fast hands the opposition an easy line about strangling a growth industry.
For now, the honest answer to the broadcaster’s question is that the policy is more likely to evolve than to lurch. The building blocks, a risk-based model, mandatory guardrails for the riskiest uses, and reliance on beefed-up existing regulators, are already visible. What remains unresolved is how firm those guardrails will be, how quickly they arrive, and whether they can satisfy a business lobby chasing certainty and a public that has learned to be wary of automated systems. That balance, more than any single announcement, is what Australians should watch.
Sources: ABC News.
