The Fox Writing the Rules

South Australia’s OpenAI deal, the Royal Commission that followed, and the question of who technology is meant to serve

I have been watching South Australia’s embrace of emerging technology for quite a few years now.

Some readers who followed my earlier work will know that I connected some of these developments on a previous platform. Smart cities, biometric identification, digital identity, energy management, emergency powers and international policy networks, which all continue to raise questions for me.

I have not entirely let go of the suspicion that there is more coordination behind these developments than has yet been shown in detail. That suspicion is still there, but I am not willing to publish it as a finding, because it isn't one yet. It is an intuition, and intuitions are exactly the thing I am asking government, industry and my own readers to be careful with in this piece. I do not get to exempt myself from that standard because the suspicion is mine.

So I want to separate two things I could easily blend together. What I privately continue to wonder about is one thing. What the evidence currently in front of me actually supports is another. This piece is built on the second, not the first.

And on the evidence, I think the narrower argument is actually the stronger one.

Different technologies can be developed for different purposes, by different institutions, with perfectly legitimate aims, and still accumulate into forms of institutional capability that previous generations did not possess. That is the question I want to consider here, and it became particularly difficult to ignore over the past few days.

Over the weekend, South Australian Premier Peter Malinauskas visited OpenAI’s headquarters in San Francisco and signed a memorandum of understanding intended to explore cooperation around AI skills, innovation, investment and infrastructure. The agreement sits alongside South Australia’s recently released Data Centre Strategy, which explicitly describes artificial intelligence as a generational economic opportunity and commits the state to actively pursuing global AI companies and data-centre investment.

Then, on Monday 10 August, the Premier announced a Royal Commission into artificial intelligence. The sequence deserves scrutiny.

This is not an argument against AI

I should make something clear at the outset: this is not a piece arguing that South Australia should reject artificial intelligence.

I use AI myself. I can see its extraordinary potential. It can assist medical research, reduce repetitive administration, improve accessibility, accelerate scientific discovery and allow a small organisation or individual to accomplish things that once required substantial resources.

Nor do I object to South Australia wanting to be innovative. A government would be negligent if it simply ignored a technology capable of transforming the economy around it.

Malinauskas himself is acknowledging both sides of that reality. He describes AI as an opportunity to advance humanity while warning that, if left unchecked, it presents what he calls a “material risk” to society and the future of work. The Royal Commission is expected to begin on 1 October 2026 and report no later than 1 July 2027. Its terms of reference are still being developed, three commissioners are yet to be appointed, and the Premier has said he wants technology companies themselves participating in the inquiry.

I do not regard that participation as inherently improper. A serious inquiry into AI should hear from the people who actually build it. But expertise and interest are not the same thing. A company may possess knowledge government desperately needs while also possessing commercial interests that government has a responsibility to scrutinise.

And when a government signs a strategic agreement with one of the world’s leading AI companies immediately before announcing an inquiry intended to help shape the rules governing that technology, I think South Australians are entitled to ask how independence will be protected.

South Australia is not starting from zero

The OpenAI agreement matters partly because it did not arrive in isolation. South Australia has deliberately positioned itself as an early adopter of digital technology for many years.

The City of Adelaide currently describes its growing network of interoperable multi-function poles as capable of carrying Wi-Fi access points and sensors. Thousands of in-ground parking sensors feed applications, networked sensors are used for waste management, and the council describes Internet-of-Things devices as allowing it to monitor the environment, public assets and how people use the city.

There are useful applications for all of these things. The point is not that a parking sensor is sinister, the point is capability.

We saw a particularly clear example during COVID. In August 2021, then premier Steven Marshall told the South Australian Parliament that the state was acting as the national pilot for a home-quarantine application using facial recognition and geolocation to remotely supervise selected people completing quarantine at home. Participants received random requests to verify their presence.

What I think can be said with confidence is: the technology has altered what government is practically capable of doing. Something that previously required physical supervision could increasingly be conducted remotely, digitally and at scale. That observation does not establish what government intends to do with similar technologies in the future. But intention is not the only question free societies should consider. Capability matters too.

Then came AI

South Australia has since moved considerably further into artificial intelligence.

The Audit Office of South Australia reported in December 2025 that an Office for AI had been established and that $28 million was allocated in the 2025–26 State Budget to an AI program across the public sector. The Audit Office found that most large agencies were still using AI cautiously, primarily through pilots and productivity tools, but some were considering limited AI decision-making. It specifically identified accountability, transparency, privacy, automation bias and public trust as governance concerns.

That seems to me an important point: the Royal Commission is not examining a hypothetical technology that South Australia might encounter one day. AI adoption inside government has already begun.

At the same time, the state’s June 2026 Data Centre Strategy says South Australia intends to pursue the physical infrastructure required to expand AI dramatically. It proposes dedicated legislation, streamlined approvals, greater powers for the Coordinator-General and strategic partnerships with global AI companies and data-centre developers.

Again, none of this establishes improper intent, it establishes direction, and it is the accumulation of these developments, rather than an allegation that they were secretly coordinated, that interests me.

The governance problem is convergence

A facial-recognition system can be built for one purpose, a digital identity system for another, a smart electricity network for a third, a government database for a fourth, an AI decision-support system for a fifth, a network of city sensors for a sixth. Each may have a reasonable justification when examined separately.

But technology does not necessarily remain separate simply because its original purposes were separate.

Increasingly, systems can talk to systems. Data can be compared with other data. Models can detect patterns across datasets that no human administrator could realistically process, and decisions can be automated, recommended or prioritised in milliseconds.

The question I am asking is not whether someone secretly planned all of this as a single architecture. I'm still discerning any evidence which could support that. I am asking what safeguards are required if separately legitimate systems gradually become capable of operating as one. That seems to me a much harder question.

Who may combine information originally collected for different purposes? When should a citizen be told that AI materially influenced a government decision about them? When must a human being remain responsible for the final decision, and what right of appeal exists when the system gets something wrong? Can information collected during an emergency later be repurposed? What happens when government becomes dependent upon a private technology provider whose system it cannot realistically replace? And which parts of ordinary life should simply remain beyond the reach of optimisation, even where optimisation is technically possible?

These are questions about limits, not motives.

The curious boundary around infrastructure

There is another reason I think the Royal Commission’s terms of reference will matter.

Malinauskas has said the Commission will consider AI’s impact across work, education, public services, health and the electricity grid. Yet he has also specifically excluded data centres themselves, saying the inquiry is about the use of AI rather than the infrastructure underpinning it.

I understand that an inquiry can become so broad that it becomes useless.

But I am not yet convinced that AI infrastructure and AI governance can be cleanly separated.

The state government’s own Data Centre Strategy says that AI is underpinned by data centres and proposes substantial regulatory and planning changes to encourage their development. It allows the Coordinator-General to exercise powers including taking over certain decision-making functions, changing assessment timeframes and streamlining regulatory processes for designated projects.

Perhaps excluding those matters from the Royal Commission is entirely sensible. But I think the government should explain why.

If we are examining the social consequences of a technology while simultaneously restructuring planning, energy and investment policy to facilitate the infrastructure on which it depends, where exactly should the boundary between the two sit?

The Christian question sits underneath all of this

For me, the deepest question is not technological, t is theological: what is a human being?

A traditional Christian understanding begins with the person created in the image and likeness of God, and that applies to everyone in this story.

Peter Malinauskas bears that image, and so do the public servants developing these policies, the engineers and executives at OpenAI, the workers whose jobs may be transformed by AI, the child encountering these systems at school, and the person whose application might eventually be assessed with the assistance of an algorithm. And so do those of us scrutinising all of them, and that matters.

Christianity does not permit me to construct a theology in which governments and corporations are fallen while critics somehow stand outside the problem. The same human tendency that can lead an institution to protect its interests can lead a writer to notice the evidence that confirms what he already suspected. I know that tendency exists in me, and it is one reason my own language around these issues remains tempered.

The answer is not therefore to find perfectly trustworthy rulers, perfectly benevolent corporations or perfectly objective critics, there aren’t any.

The Christian answer to fallenness has always included limits, accountability, repentance and the refusal to place ultimate trust in human power. That is where I think Christianity has something important to contribute to the AI debate.

Efficiency is not the highest good

Artificial intelligence is very good at something governments value enormously: reducing friction.

It can process applications faster, detect patterns earlier, allocate resources more efficiently, predict risks, reduce costs, identify anomalies and automate decisions.

Much of that may be beneficial. But there is a danger in allowing the logic of the tool to become the logic of the society using it, human beings themselves create friction.

We disagree.

We change our minds.

We forgive.

We make exceptions.

We act from conscience rather than probability.

We care for people who may never produce an economic return on the resources invested in them.

We sometimes refuse the most efficient option because we believe it is morally wrong.

From a Christian perspective, these are not necessarily bugs in the human system, sometimes they are features of being human. A person is not merely a unit of productivity, a behavioural profile, a risk category or a collection of data points.

Human worth does not increase because an individual is useful to the state or profitable to the market, nor does it diminish because somebody is inefficient, dependent, disabled, elderly, poor, unborn, unemployed or difficult to manage.

That is where the Imago Dei becomes politically inconvenient. It places a limit around every system.

The fox and the rules

Which brings me back to the deliberately provocative title. I am not claiming OpenAI is writing South Australian law, there is no evidence before me that establishes that. The fox metaphor points instead to an old governance problem. Those possessing expertise will naturally influence the rules governing their area of expertise. Governments need their knowledge. Industry wants workable regulation. Researchers understand technical possibilities that politicians cannot reasonably be expected to master.

But the parties being regulated also have interests. South Australia has entered into an agreement with OpenAI while simultaneously seeking major AI investment. The government is developing legislation intended to facilitate AI infrastructure, and it now proposes a Royal Commission whose recommendations may influence how that technology is governed. Malinauskas has said he hopes AI companies themselves will participate and that successful recommendations may eventually be followed by other states, none of that proves regulatory capture.

It does create conditions in which guarding against regulatory capture matters.

So I want to know whether the OpenAI agreement will be publicly available in sufficient detail for South Australians to understand it, what relationship, if any, exists between that agreement and the Royal Commission, and whether commissioners will disclose relevant industry relationships.

I want to know whether civil-liberties organisations, families, churches, small businesses and ordinary citizens will have meaningful opportunities to be heard alongside technology companies, unions, universities and government agencies.

I want to know whether human review, appeal rights, purpose limitation, data minimisation and meaningful non-digital alternatives will form part of the discussion.

And I want to know whether the inquiry will examine the possibility that something can be technologically impressive, economically profitable and administratively efficient while still diminishing something properly human.

Preserving the person who can still say no

I once spoke about protecting the “independent actor” largely in terms of remaining outside systems. I think I would describe it differently now. The independent actor does not need to abandon technology. He can use AI. She can carry a smartphone. A family can live in a smart city. A household can participate in a modern electricity network.

The deeper question is whether people retain meaningful moral agency while doing so.

Can a citizen challenge an automated decision? Can a parent still exercise genuine authority over the formation of a child? Can a church order its life according to its faith, and can local associations form without needing administrative permission for every meaningful activity? Can somebody request human judgment when an algorithm has categorised them incorrectly, and can a person decline a particular technology without becoming practically excluded from ordinary civic life? Can we still say no?

What unsettles me more is not a question I can fully answer. Is it the possibility that no one is in control of where this goes, that capability is simply accumulating because that is what interoperable systems do? Or is it the possibility that someone is, and has reasons they have not shared? I do not know which of those two thoughts troubles me more, and I am not going to pretend I have settled it.

That capacity matters because freedom is not simply the existence of choices inside a system. Sometimes freedom requires being able to challenge the system itself. And sometimes what looks like friction from the centre is liberty from the ground.

What would leadership look like?

South Australia may genuinely have an opportunity to lead. Perhaps the OpenAI agreement will prove beneficial. Perhaps the Royal Commission will become one of the strongest examinations of artificial intelligence yet undertaken in Australia. I hope so.

But I am asking whether leadership in this area should be measured by something more than how quickly we can attract investment or deploy new capabilities. Could technological leadership also mean being unusually serious about restraint?

Could South Australia insist that government AI decisions remain explainable and appealable, and could it build genuine separation between commercial partnership and regulatory influence?

Could it protect reasonable avenues for human interaction as services become increasingly automated, and ensure that data collected for one legitimate purpose does not simply become available for every future purpose?

Could it ask not merely whether something can be optimised, but whether it should be?

And could the Royal Commission begin not with the economic question, “What can AI do for South Australia?”, but with the older and more important question:

What is a human being, and what may no institution legitimately do to one?

For Christians, Christ gives us a measure of authority very different from domination. He tells His disciples that worldly rulers “lord it over” others, before saying:

“Yet it shall not be so among you; but whoever desires to become great among you shall be your servant.” — Mark 10:43

That principle reaches well beyond artificial intelligence.

Authority is legitimate when it serves.

Technology is good when it serves.

Markets are useful when they serve.

Government has a proper place when it serves.

None of them is the measure of the human person.

The person comes first.

So as South Australia races toward an AI future, I am not asking that we turn around. I am asking whether we are building the guardrails while we can still decide where they belong.

And, as always, if I have drawn a connection the evidence does not support, or missed something that materially changes this account, I welcome the correction. The question is too important to require anyone to pretend they have already arrived at the answer.

Thanks for reading.

God bless.

Mark.