South Australia Just Funded Australia's First AI Royal Commission. The Rules That Last Will Be the Ones About People.

South Australia has committed AU$3 million and 3 royal commissioners to the first major government inquiry into artificial intelligence in Australia. Premier Peter Malinauskas announced it on August 10, 2026. Hearings open in October, the final report is due in July 2027, and the terms of reference arrive within 4 to 6 weeks. So the question worth asking now, while those terms are still being drafted, is which of the inquiry's findings will still be true on the day the report is printed.

Talk to any working illustrator about the last 2 years and you get the version of this story that never fits neatly onto a submission form. Same client. Same brief. A smaller fee, and a line in the email about what the model had already produced. That person is on the evidence list for this inquiry, which is the most interesting decision South Australia has made so far.

What South Australia actually built

The scope covers employment and the future of work, education, industry, public health, and the arts, and evidence will be taken from business, industry, unions, technology developers, academics, and the creative sector. Credit where it belongs: an inquiry that hears only from the people building the technology produces a different report than one that also hears from the people whose work it reprices. South Australia is not approaching this cold, either. Malinauskas met representatives of OpenAI, Anthropic, and Apple during a recent visit to the United States, and the state had already signed a memorandum of understanding with OpenAI. It engaged the industry directly, then decided the public deserved a formal record.

"We are at a really important fork in the road. We can choose to allow this industry to come to our country and evolve in a way that is unregulated. Or we can choose to set the agenda … to ensure that we use AI in a way that puts people first."

Peter Malinauskas, Premier of South Australia, August 10, 2026

Setting the agenda is the right ambition, and it raises a design question every rule-making body in the world is working under right now. Deliberative process runs on a calendar because it has to: hearings get scheduled, submissions close, and findings are tested by people who can be cross-examined before anyone publishes them. That discipline is what makes a royal commission's conclusions worth more than a press release. Model capability keeps a different schedule entirely. No government invented that arrangement and none of them can vote it away; it is the arithmetic of governing a technology whose capability curve refuses to sit still.

The rules that outlive the model

Which is why the most durable thing this inquiry can write down is a set of obligations rather than a description of what the technology could do in 2027. Who has to disclose what. Who answers when an automated decision turns out to be wrong. What an employer owes a worker whose job is redesigned in the middle of her career. Those questions have the same answer in October 2026, in July 2027, and in 2031.

Two live demonstrations are already in the record. A United States appeals court ruled this month that a computer-crime statute from 1986, written before the commercial web existed, decides who is responsible when an AI shopping agent acts on your behalf; the law survived because it describes conduct and responsibility rather than any particular technology. Europe's requirement that a chatbot tell you it is a machine became enforceable this month, and it holds regardless of how capable that machine becomes. Rules written about capability expire with the model. Rules written about accountability outlive it.

Governments have moved unusually fast when they decided a moment demanded it. Sputnik went up in October 1957, and a national education act was on the books inside a year. The money went to teachers, laboratories, and language programs, which is to say that it went to people. A generation of students who had no idea what a satellite had to do with their chemistry class finished school with a different set of skills because of it. Speed was possible because the decision was about people, and a decision about people never has to wait for the technology to settle.

There is real work the rest of us can do while the commissioners do theirs, and roughly a month to prepare it. Employers can bring the specific role that changed, what the retraining cost, and how long it took. Educators can bring the date a syllabus went stale and what replacing it required. Unions can put a name and an age to the worker whose job was rebuilt around a tool nobody had trained her to use. Evidence in that shape can be acted on inside a formal process, and opinions in submission form mostly cannot. I advise the California State University system on AI governance and hold an innovator-in-residence appointment at Tulane, and the pattern underneath most AI governance work is a deferred bill: every year retraining is postponed, the learning debt compounds, and the interest is paid by people in the middle of their careers.

That scope list reads like a policy document, and it will settle very ordinary things. Whether the 52-year-old warehouse supervisor gets a bridge into her next decade or a redundancy letter. Whether a nursing student's clinical training still matches the ward she walks into after graduation. South Australia has given itself 9 months of hearings to get answers like those onto a public record the whole country can read.

Three commissioners cannot resolve all of that, and no inquiry was ever built to. What this one can do is set the terms everybody else argues from for the next decade, and the terms that hold are the ones written about people. So here is the question for your own next leadership meeting: which of your AI rules would still be true if the model you use doubled in capability tomorrow, and which of them were really written about one particular tool? Our terms of reference are shorter than a commission's, and we get to revise them far more often. If you are preparing a submission, or writing the rules your own organization will live by, I am easy to find.

Sources: SBS News, August 10, 2026 · News24 Australia, August 10, 2026 · The Mandarin, August 10, 2026.