ACT eyes NSW take-home ban as moratorium contagion spreads

Digital Attitudes, 31 August 2026

Sunday, and the moratorium NSW ordered a fortnight ago is now openly under review one border across, at the same time a US research group is publishing what the third way between ban and free use of classroom AI could actually look like.

ACT considers the NSW take-home ban despite a 0.1 per cent breach rate

The ACT government confirmed to the Canberra Times on Saturday 23 August that it is actively reviewing the NSW decision to ask NESA for a moratorium on unsupervised take-home assessment tasks, with the Board of Senior Secondary Studies "continually monitoring the impact of AI in schools" (Canberra Times on the ACT reviewing NSW take-home ban). The territory recorded about 0.1 per cent of assessments with academic integrity breaches over the past year, an order of magnitude below the figures NSW cited when Prue Car wrote to NESA chair Peter Shergold on 31 July, and Dr Luke Rowe of Australian Catholic University still backs the NSW position because "you've got to have circumstances that allow you to isolate the student's capacity away from the technology."

Why this matters: For any Australian Deputy Principal, Head of Curriculum, or senior secondary studies coordinator watching a jurisdiction the size of NSW take a specific assessment tool off the table, this is the moment the contagion becomes visible. The ACT's breach data does not remotely justify a ban on its own terms, but the political calculus of being the outlier when NSW moves is different to the pedagogical one. Term 4 planning for the Year 12 class of 2027 should include a scenario where Victoria, Queensland and South Australia read the NESA advice and, one by one, ask their own regulators what a defensible position looks like. The moratorium is now a national conversation whether or not it ever becomes national policy.

NC State's Socratic Challenger interrogates the student, does not write for them

An interdisciplinary team from North Carolina State University published a proof of concept on Tuesday 26 August for what they call the Socratic Challenger, an eight-step research question workflow that uses generative AI at five specific steps to push back on a student's thinking rather than generate content on their behalf (North Carolina State on the Socratic Challenger framework). Associate Professor Aram Mikaelyan frames it as "the AI is essentially being used to interrogate the student's process", and a nine-week ecology cohort of 45 undergraduates produced research questions the instructors described as substantively better without the AI drafting any writing for them.

Why this matters: For any Australian university Head of Learning Design, Dean of Teaching, or course coordinator drafting a 2027 AI use policy that has to sit somewhere between USyd's full disclosure position and UNSW's Level 1 default prohibition, this is a public workflow with published results to point to. The pedagogy is the deliverable, not the tool. Every Term 4 policy conversation currently getting stuck on "how do we detect it" can now be reframed as "how do we structure the assessment so the AI works against the student's laziness rather than for it". For NESA and TEQSA writing sector guidance on assessment reform, the framework is a live, testable answer to what the third pathway between ban and free use actually looks like inside a syllabus.

OpenAI adds 55 US districts to free ChatGPT for Teachers, no Australia yet

OpenAI announced on Tuesday 26 August that ChatGPT for Teachers is expanding to 55 additional US school systems across 20 states, adding more than 100,000 educators and bringing the total to over 300,000 teachers across 100 K to 12 organisations in 30 states, with the free tier locked in until June 2028 (9to5Mac on the free ChatGPT for Teachers expansion). Alongside the district additions, OpenAI published a shared data privacy agreement covering 16 US states, letting districts evaluate the product against a single framework rather than 50 separate legal reads.

Why this matters: For Australian principals and IT leaders who have been waiting for the OpenAI for Australia launch of late 2025 to translate into a teacher tier, this is the fourth US expansion in as many months and there is still no equivalent Australian offering. NSW public schools have NSWEduChat, South Australia has EdChat, Queensland has Corella, but no Australian jurisdiction has a free, direct ChatGPT for Teachers arrangement with a shared privacy schedule. The 16 state data privacy agreement is the piece Federation of Australian Government Schools procurement leads and the OAIC should be reading closely, because a pre-agreed, cross-jurisdictional privacy schedule is exactly the shape the Commonwealth Digital Duty of Care work is heading toward. If OpenAI is going to sell this pattern into Australia, it will land first as a national schedule, not as one contract per state.

Three jurisdictions in three different postures on classroom AI, one research group pointing at how it should actually work, and one week where the pedagogy conversation stopped being about detection and started being about design. The full operational read is at digitalattitudes.com.au.