Only one ANZ university has a dedicated AI policy

Digital Attitudes, 12 September 2026

Saturday, and two weeks after Prue Car's take-home cap and a week after UNSW's ChatGPT Edu rollout, a Heliyon paper counts what 49 Australian and New Zealand universities have actually written down about generative AI. The answer is that most of the policy work is still not on paper.

One in 49 ANZ universities has a dedicated AI policy, and it is UTS

The paper by Dadich and Dhakal, published this week in Heliyon, audits public GenAI policies at 49 universities across Australia and New Zealand and finds only 22 explicitly mention generative AI, only one (the University of Technology Sydney) has a dedicated AI policy, and across the corpus the word "integrity" barely registers (Bioengineer's coverage of the Dadich and Dhakal Heliyon paper). Scholarly output on the same topic went from 15 papers in 2023 to 76 in 2024, so the sector is talking about GenAI at five times the rate it is writing rules for it. The authors call the dominant institutional stance "moderative": use GenAI when a lecturer says you can, otherwise assume you cannot.

Why this matters: TEQSA's regulatory-led posture for 2026 asks universities to show how they manage AI risk against the Higher Education Standards. Twenty-seven ANZ institutions cannot show that with a policy that does not name the thing. If the audit lands on the DVC (Academic) desk on Monday, the practical question takes about a paragraph to answer: what does our AI policy actually say a student must do, and where does it live where a student will actually find it. If the answer is "buried in an academic integrity procedure last updated in 2022", that is the fix. UTS has a template. The paper effectively points at it.

Boston bakes AI fluency into a high school diploma this month

Boston Public Schools starts the 2026-27 school year with AI literacy as a graduation requirement across every high school in the district, a first for a major American city, funded by a $1 million seed grant from Kayak co-founder Paul English and a curriculum built with UMass Boston (GovTech on the Boston Public Schools rollout). The framing is skills-first: students leave with the capacity to use AI in college and work, not the option to avoid it. It sits at the opposite end of the same fortnight in which New York City's mayor pulled generative AI from K to 8 public schools for the full year (covered here on 7 September).

Why this matters: Australian schools are being asked the same question in slower motion. NESA capped take-home HSC assessments on 1 September, NSW EduChat opened to Year 5 to Year 12 students last month, and the AITSL teacher survey covered on Friday did not ask a single AI question of 50,000 respondents. Three American city districts (Boston, New York, Katy) have picked three different answers to whether AI belongs in the graduation profile at all, and they have all committed the answer to writing. Australian principals writing their 2027 Strategic Improvement Plans this term have to pick one and put it in writing, because "the sector is still figuring it out" will not survive a term of parent questions about the HSC cap.

Character.AI pulls teen chat, Chub geo-blocks Australia, and the codes are working

The eSafety Age-Restricted Codes have been enforceable since 9 March 2026, civil penalties reach $49.5 million per breach, and the industry response has landed. Character.AI removed open-ended chat for under 18s (rolled out from 24 November 2025), Chub.AI withdrew from the Australian market entirely, Chai moved companion chat behind a paid subscription, and Nomi committed to further age assurance (Commsroom on how the eSafety codes now cover AI chatbots, Character.AI's own note on the under 18 changes). This is the practical outcome of the same regulatory arm Wells named in Tuesday's Digital Duty of Care draft, six months earlier and on a narrower slice of the market.

Why this matters: Every school Acceptable Use Policy written before March 2026 has language about "AI companion apps" that assumed those apps behaved the same way for a 14 year old as for a 40 year old. That assumption is not true anymore, and the compliance surface has already moved. On a BYOD device, Character.AI's under 18 experience is now non-chat only, Chub is gone from the App Store here, and the IT team's block list should shrink not grow. The behaviour management conversation is different too: a Year 9 who claims a Character.AI "girlfriend" told them to skip class is now, on the vendor's own admission, using a non-chat feature, and the class debrief runs differently. Update the AUP once, tell parents in the next newsletter, and move on.

The three stories are one story. What a system writes down about AI is now downstream of what the regulator, the vendor, or the graduation profile has already committed to paper. If your institution's policy is still verbal, this weekend is a decent time to fix that. More on our view of what a good school AI policy looks like at digitalattitudes.com.au.