South Australia’s AI royal commission faces a scope test before it starts

The costs are also not yet clear in public accounts. Malinauskas told ABC News he expected a cost of about $3 million; another report said he acknowledged a cost well above $3 million. The appointments and final remit will put a firmer boundary around that commitment.

An AI-chip illustration credited to Igor Omilaev on Unsplash. Source: ABC News.

The data-centre question has not gone away

Malinauskas has drawn a distinction between the use of AI and the data centres that enable it. He said the royal commission would not examine that infrastructure, arguing that how the technology is used in society is more consequential than where a data centre is built, at the announcement.

But the government’s preliminary topics expressly include AI-related infrastructure and the consequences for energy, water and the grid. That is not necessarily a contradiction: the final terms could distinguish individual development approvals from the system effects of AI infrastructure. Until they are published, it is not possible to tell where the inquiry will place that boundary.

The distinction matters because the state is simultaneously seeking data-centre investment. It released a data-centre strategy in June, and Iren, a Nasdaq-listed company founded by Australian brothers Daniel and Will Roberts, announced a $10 billion data-centre project at Bundey, 165 kilometres north-east of Adelaide, according to the report. The announced amount describes the project, not a finding that the investment will deliver its claimed benefits.

Federal policy already proposes a separate route for those constraints. A July government release says large data centres would be required to underwrite new power supply, pay their full connection costs, reduce demand when needed for the grid and improve water efficiency. Its approach was to go to National Cabinet in August, with legislation expected early the following year, the federal release says.

Deployment choices are nearer than a national rulebook

The commission’s direct practical choices lie in areas the state controls or can influence: public-service use, education, workforce planning and policy recommendations. Malinauskas said South Australia had not adopted the outright ChatGPT bans used in some other states; instead, he argued schools should teach children to use AI in a regulated way. He also said students should not leave school unable to calculate without it, in outlining the proposal.

For regulated professions, Laina Chan, a practising barrister and the founder and chief executive of MiAI Law, a legal-AI research and risk-management system, said the inquiry should distinguish situations where AI can safely augment expertise from those where human professional judgment must remain central. Her concern is not that professional knowledge is simply stored information: lawyers and doctors develop judgment through education, supervised practice and accountability, she said, in her response.

There is also a competing economic view. Anton van den Hengel, chief scientist at the Australian Institute for Machine Learning, said South Australia’s research and startup ecosystem gives it a basis to participate in AI’s upside. He warned that making it too difficult for AI companies to prosper locally could leave Australia importing other countries’ systems instead of shaping their development, in an interview about the plan.

Andrew Hastie, the federal opposition’s shadow industry minister, has proposed an alternative institutional response: a tripartisan federal parliamentary committee. He said AI’s strategic, moral and economic implications made it a national-parliament question, in his response.

What must be decided before October

The next evidence is procedural, not technological: the government must publish the terms of reference and name the three commissioners. Those documents will show whether the commission is concentrating on a defined set of deployments and state responsibilities, or trying to cover every social and infrastructure consequence of AI on a short clock.

They will also show how its work relates to the federal standards process. A commission that can identify concrete choices on schools, public procurement, skills, professional safeguards and infrastructure effects could add useful evidence to broader policy. If its remit remains expansive or overlaps indistinctly with data-centre decisions and national regulation, its July 2027 report may arrive with its sharpest choices still unresolved.