Australia Is Now Formally Worried About How We Think: What the AI Inquiries Mean for Learning

By Michael Peart, Managing Partner, Hungry Minds Learning

Key takeaways

• South Australia announced Australia’s first royal commission into AI on 10 August 2026, covering work, education, public services and the creative industries. Commissioners start in October 2026 and report by 1 July 2027.
• On the same day, NSW asked NESA to consider a moratorium on unsupervised take-home HSC assessment because of AI.
• A federal Senate inquiry into AI and data centres, referred in May 2026, reports on 16 November 2026.
• Premier Peter Malinauskas’ stated worry is students leaving school fluent in AI tools but unable to think deeply or communicate with each other.
• For employers this is a workforce pipeline issue, not an education issue. The graduates arriving in 2028 will be technically capable and conversationally underdone.

Three separate Australian processes are now formally examining what AI is doing to how people think, and two of them landed in the same week. If you run learning and development, this is not background noise. It is the beginning of a policy environment that will shape what you are expected to prove about capability.

What is underway right now

On 10 August 2026, South Australian Premier Peter Malinauskas announced Australia’s first royal commission into artificial intelligence. It will examine AI’s effect on work, education, public services and the creative industries. Terms of reference are due within four to six weeks, three commissioners will be recruited internationally and start in October, and the final report is due by 1 July 2027. The cost is around $3 million, which the Premier described as trivial against the questions involved. The state Liberal leader, Ashton Hurn, has questioned whether it duplicates a parliamentary committee that already made 15 recommendations, and federal independent MP Kate Chaney has argued a royal commission is too slow a vehicle for a technology moving this fast.

The announcement followed a US trip where Malinauskas met OpenAI, Anthropic and Apple, signing a memorandum of understanding with OpenAI president Greg Brockman and meeting Anthropic co-founder Tom Brown. The federal government signed its own MoU with Anthropic in April 2026, as a commitment under the National AI Plan released in December 2025.

On the same day in NSW, Deputy Premier and Education Minister Prue Car asked NESA to consider a moratorium on unsupervised take-home assessment tasks for the HSC, pending a review of AI’s effect on student learning. Half of an HSC result comes from school-based assessment, some of it done at home.

Federally, the Senate Environment and Communications References Committee is running an inquiry into artificial intelligence and data centres, referred in May 2026, and it reports on 16 November 2026. That sits on top of the earlier Senate Select Committee on Adopting AI, whose final report the government formally responded to in April 2026. So: a royal commission in Adelaide, an assessment moratorium in Sydney, and a live committee process in Canberra.inquiry into artificial intelligence and data centresSenate Select Committee on Adopting AI

The concern that should interest L&D

Most coverage of the SA royal commission has focused on data centres, energy and jobs. The line that stopped me was about education. Malinauskas said he hates the idea of “kids graduating from high school and they’ve learned how to use ChatGPT” but not how to read, write, do arithmetic or communicate with one another. He also said he considers AI more consequential than social media, which is a striking claim from the government that led the under-16s ban.

He is describing a specific failure mode, and it is not “kids are lazy”. It is that fluency with a tool can look identical to competence at the underlying task, right up to the moment someone has to perform without it. A student who can prompt their way to a distinction essay has demonstrated real skill. What they have not necessarily demonstrated is the ability to hold a complex argument in their head, sit with an unresolved problem, or negotiate meaning with another human being in real time.

There is now evidence for this rather than just intuition. The MIT Media Lab’s cognitive debt research (Kosmyna et al., Your Brain on ChatGPT, 2025) found that participants who wrote essays with ChatGPT showed the weakest neural connectivity of any group — up to 55 per cent lower than those writing unaided — and that 83 per cent were unable to quote from the essay they had just written. It is a preprint rather than a peer-reviewed paper, the sample was small at 18 people per group, and other researchers have questioned the methodology — so treat it as a signal rather than settled science. I have written about what that means for workplace training in Completed the Course, Learned Nothing, including why AI detection tools cannot rescue the situation and what has to change in design instead.

What I am doing about it as a parent

I have 16-year-old twins, so I am not observing this from a distance. My view is that you cannot solve this by restricting the technology, and I would not want to. AI fluency will be non-negotiable in their working lives. What you can do is deliberately build the capabilities AI does not build, and most of those are built by other people rather than by screens.

So I insist on part-time or casual work, and I am specific about the setting. I push them towards busy hospitality: pubs and restaurants where they work both the kitchen and the dining floor. That combination is not accidental. The kitchen teaches sequencing under time pressure, taking direction, and holding your part of a system that fails visibly if you drop it. The floor teaches reading a stranger’s mood in three seconds, de-escalating someone who is unhappy, and making conversation with people who share nothing with you. There is no prompt for either. A Friday night service is a better communication curriculum than anything I could design for them.

I also want them entrepreneurial, and genuinely good with AI rather than nervously avoiding it. Both work casually at Hungry Minds, where their job is to help build the challenging tasks and questions we put into eLearning. The rationale is partly selfish: teenagers are ruthless at spotting a question you can answer without thinking, and ruthless at spotting content that is boring. If a scenario survives their scrutiny, it will probably survive a learner’s.

The unexpected benefit has been what it does to our senior consultants. Having sixteen-year-olds in the room asking why a task is written that way keeps very experienced designers fresh and slightly on edge, in the good sense. It is a two-way apprenticeship: the kids get to talk with seasoned professionals about real work, and the professionals get their assumptions poked by someone with no investment in how it has always been done. It is also, straightforwardly, fun.

Why employers should care now

The royal commission reports in July 2027. The NSW changes could hit classrooms by Term 4 this year. Neither of those timelines matters as much as the simple arithmetic: the cohort currently in Year 11 enters your workforce around 2028. They will arrive more technically capable than any previous graduate intake and, on the evidence and the Premier’s reading of it, less practised at unaided reasoning and face-to-face communication.

That is an onboarding and capability problem, and it lands on L&D. It means induction and early-career development cannot assume the communication and critical thinking baseline we used to assume. It means the learning campaign model matters more than ever, because capability built through spaced, applied, multi-touchpoint reinforcement is the only kind that survives. And it means blended and face-to-face design is not a nostalgic preference. Time in a room with other humans is now a scarce developmental input.

Practical ways to stay ahead

Split into two lists, because the parenting version and the organisational version are different problems.

If you are raising teenagers

1. Insist on customer-facing casual work. Hospitality, retail and hands-on service roles force real-time human negotiation under pressure. Prioritise settings where they must work in a team and speak to strangers, not roles where they sit alone with a screen.
2. Do not ban AI, supervise it. Sit with them while they use it. Ask what they asked it, what it got wrong, and how they checked. The skill you are building is scepticism, not abstinence.
3. Make them do the thinking first. The household rule that matters most: form your own view, in your own words, before you open the model. Then use AI to challenge and improve it.
4. Protect handwriting and reading. Longhand notes and physical books are slower on purpose. That friction is where the processing happens.
5. Give them real work with real stakes. A casual job inside a business, a small enterprise of their own, anything where the output is used by someone who will complain if it is wrong. Consequence teaches faster than assessment.
6. Put them in rooms with adults. Deliberately create situations where they talk with experienced professionals about actual problems. Confidence with adults is a learnable skill and school rarely teaches it.

If you run L&D or a business

1. Stop treating completions as evidence. A finished module with a passed quiz now tells you almost nothing. Shift reporting to behaviour change, on-the-job application and manager-verified capability.
2. Move assessment towards demonstration. Observed tasks, scenario walk-throughs, verbal explanation, workplace evidence. If a chatbot can pass it, it was measuring recognition rather than capability.
3. Design campaigns, not events. One module or one workshop day cannot hold against either the forgetting curve or AI shortcutting. Use spaced touchpoints across weeks with retrieval built in.
4. Rebuild the manager conversation. Equip team leaders with structured toolkits and extension prompts so they can probe understanding live. A five-minute discussion verifies more than a ten-question quiz.
5. Invest in face-to-face deliberately. Treat in-person time as the place where communication, judgement and disagreement get practised, and protect it with device agreements and scheduled digital breaks rather than letting laptops eat the room.
6. Teach AI literacy as a capability. Most employees have had no instruction in the difference between AI as a coach and AI as a ghostwriter. Make it explicit, including where its use is legitimate and expected.
7. Bring young people into your design process. Teenagers and early-career staff will find the weak questions in your courses faster than a QA checklist, and the exposure works in both directions.
8. Put your view on the record. The SA royal commission’s terms of reference were due within four to six weeks of the August announcement, and the commissioners start in October. If AI’s effect on capability matters to your sector, this is the window where the framing gets set, and an industry body is usually the fastest way in.

The wider point

Three governments have now decided that AI’s effect on human thinking is a policy problem worth formal machinery. Whether a royal commission is the right instrument is a fair argument, and the critics have a point about speed. But the underlying observation is sound, and it is one L&D has been circling for years without the language for it: we have become very good at recording that learning happened and increasingly poor at knowing whether it did.

Inquiries will not fix that. Better design will. If you want the detail on what that looks like in practice, the companion piece to this article sets out ten specific techniques for building eLearning and face-to-face training that AI cannot complete on the learner’s behalf: Completed the Course, Learned Nothing: AI and the Quiet Erosion of Workplace Learning. Or see how we build multi-touchpoint learning campaigns, eLearning that holds attention and job aids that support application on the job.