UNSW opens ChatGPT to 80,000, AITSL survey skips AI

Digital Attitudes, 11 September 2026

Friday of Term 3 Week 9, and while the sector is still working through Wednesday's Digital Duty of Care exposure draft and yesterday's Canvas Lite launch, UNSW has quietly signed the largest ChatGPT Edu deployment in Asia Pacific with a three-lane assessment rulebook, AITSL has published its biggest ever teacher survey with no AI question in it, and CSER has opened a free AI teaching program to every Australian school.

UNSW hands ChatGPT to 80,000, and the assessment rules come in three lanes

UNSW Sydney announced on Wednesday 9 September 2026 that its strategic partnership with OpenAI will bring ChatGPT Edu to more than 80,000 students, staff and researchers, with Higher Degree Research candidates and casual academic staff gaining access this week and full student rollout targeted for the 2027 academic year, the largest ChatGPT education deployment in the Asia Pacific region and OpenAI's first strategic education collaboration in Australia (UNSW's own announcement of the partnership with OpenAI, iTnews on the new UNSW measures to deal with AI cheating, IT Brief on the UNSW ChatGPT Edu launch for 80,000). The rollout ships alongside an "assessment transformation" with three explicit AI use categories. Can (permitted within course guidelines), Can't (prohibited when independent knowledge or skill demonstration is required), and Must (required to demonstrate AI literacy and discipline capability). UNSW's own line on detection: "detection tools and watermarking will not be relied upon in isolation for determining misconduct. They represent a tool to support educator judgement". Deputy Vice-Chancellor Education and Student Experience Sarah Maddison, on the record: "Our ambition is to produce AI-capable graduates". OpenAI's Asia Pacific Managing Director Kiran Mani, in the same release: "Australian students want to do more than learn about AI. They want to use it to build companies, advance research and solve real-world problems".

Why this matters: for a pro vice-chancellor academic, director of digital learning, or head of a school of computing sitting inside the Group of Eight or ATN reading the USyd picket line and the ANU CIO appointment in the same term, this is the vendor floor moving. Two operational reads. First, the assessment framework. Can, Can't, Must is the first Australian university rulebook that names AI use as a required competency in some tasks rather than a policy to be worked around, and it will now be read against by every 2027 course accreditation and every enterprise agreement clause landing before Term 4. The USyd bargaining ask (2 September brief) was about enforceable AI protections. UNSW has just answered the opposite half of that question. What does the university expect an AI-capable graduate to do, and where in the curriculum is that measured. A faculty board that cannot answer both questions by November will be answering them under a UNSW comparison. Second, the vendor route. OpenAI has now signed its first strategic Australian university partnership at UNSW in the same term Anthropic ran its Claude Campus ambassador program (applications closed last night, Tuesday brief) and holds the AUD 3 million MOU that put Claude into ANU computer science courses. The 2027 Australian university AI stack is being carved between two US frontier vendors in the same term the Digital Duty of Care exposure draft names AI chatbots for the first time (Wednesday brief). If your 2027 procurement paper does not answer "which frontier vendor is this student on, and which enterprise agreement clause governs their staff use", it is not a procurement paper.

AITSL's biggest teacher survey has a 50,000 teacher blind spot on AI

AITSL chief executive Tim Bullard sat down with The Educator this week to unpack the National Trends 2026 read of the Australian Teacher Workforce Survey, drawing on responses from nearly 50,000 teachers, and confirmed in the piece published on Thursday 10 September 2026 that the June 2026 publication "does not specifically capture teachers' views, confidence, or professional learning needs in relation to AI" (The Educator K/12 deep dive on the biggest teacher survey and what it means for 2027). AI appears in the National Trends 2026 report only in the context of the federal Workload Reduction Fund pilots covering administrative tasks and marking assistance, not as survey data on actual teacher AI experiences. Bullard, on the record: "AI is certainly an immediate area of rapid change and immediate focus". And separately: "AI literacy is not simply about knowing how to use a particular tool". And finally, that any teaching adoption of AI has to be judged on whether it "will amplify or detract from quality teaching and learning".

Why this matters: for a school business manager, director of digital learning, or head of teaching and learning building a 2027 workforce plan on top of last week's ReadyTech benchmark (67 per cent of Australian education providers still without a documented AI policy, Sunday brief) and Monday's Instructure ANZ finding (28 per cent of educators feel their institution has a clear AI strategy), the sector's most authoritative teacher data set has just confirmed it did not measure teacher AI use. Two operational reads. First, the workforce planning hole. Any professional development plan citing National Trends 2026 for its "gap analysis" on AI capability is citing a survey that did not ask the question. A school leadership team writing a 2027 professional learning plan that starts with "AITSL data shows" is starting from a blank page it did not know was blank, and the CFO signing off the PD budget is now signing off against a benchmark that is not there. Second, the vendor conversation. UNSW ships an "assessment transformation" today (see above) and Corella, NSWEduChat, EdChat and ClassmAIte are all sitting in schools whose staff AI use has never been formally measured at scale by the national teacher regulator. Every vendor slide from Term 4 forward that cites "teacher readiness" now has to answer for whose data. The AITSL 2027 survey design conversation is the one to walk into now, not the one to read the results of in June 2028.

Adelaide University and Google.org open a free AI teaching lane across the curriculum

The Computer Science Education Research Group at Adelaide University announced on Wednesday 2 September 2026 that it has partnered with the UK-based Raspberry Pi Foundation to bring Experience AI, the AI teaching program developed with Google DeepMind and funded in Australia by Google.org, to Australian classrooms, with free lesson plans, slide decks, videos, student activities, assessments and six hours of professional learning at no cost to teachers, aligned to Australian Curriculum Version 9.0 from Foundation to Year 10 and to the Australian Institute for Teaching and School Leadership standards (The Educator on the global AI teaching program arriving in Australian classrooms, AZoRobotics on the new CSER partnership teaching AI to educators, Raspberry Pi Foundation on supporting AI education for 150,000 learners in Aotearoa New Zealand and Australia). CSER Project Lead Dr Rebecca Vivian, in the release: "Teachers will have access to lesson plans with speaker notes, slide decks, videos, student activities, and assessments". Raspberry Pi Foundation chief executive Philip Colligan, in the same materials: "We are looking forward to equipping teachers with the skills needed to teach AI confidently". The program reaches non-specialist teachers across every subject with no prior technical knowledge required, and covers machine learning fundamentals plus ethics, bias, deepfakes and responsible generative AI use. Experience AI has reached roughly 1.2 million learners across 160 countries since February 2025.

Why this matters: for a Foundation to Year 10 curriculum coordinator, a K to 12 digital learning lead in a Catholic or Independent tenant, or a state education department procurement officer weighing an AI literacy contract for 2027, the shape of the AI literacy market just changed. Two operational reads. First, the price floor. The CSER-cited numbers are that 66 per cent of Australian lower secondary teachers are already using AI in the classroom, only 32 per cent have access to their institution's approved tools, and 96 per cent said they would accept free professional learning if it were offered. A free, Curriculum V9 aligned program run through the Raspberry Pi Foundation and Adelaide University now sits on the Term 4 procurement panel next to any commercial AI literacy vendor that has been pitching a per-student licence, and that vendor now has to answer "why not free". Second, the sponsorship read. Google.org is funding the Australian delivery of a program built by Google DeepMind, and the same week UNSW committed 80,000 users to OpenAI. Corella, NSWEduChat, EdChat and ClassmAIte have all been positioned as sovereign alternatives to the hyperscaler stack, yet the free AI literacy lane into every Australian classroom is now underwritten by Alphabet's philanthropic arm in an academic partnership with a public Australian university. Any school council paper naming a "sovereign AI strategy" in September needs a paragraph on what happens when the free option in the room comes from one of the two US frontier vendors.

Three moves, one Friday. UNSW has signed the largest ChatGPT Edu deployment in Asia Pacific and written the first Can, Can't, Must assessment rulebook to go with it. The national teacher regulator has confirmed its biggest workforce survey did not measure AI. And a Google-funded, Adelaide-delivered free AI literacy program has opened the price floor for every Australian classroom in the same week. For a longer read on how to think about the vendor floor before Term 4 procurement papers close, see digitalattitudes.com.au.