AI in Schools
NAPLAN writing data lands at schools, Cogniti goes global
Digital Attitudes, 17 June 2026
The Wednesday read is one assessment platform releasing the writing data the year has been waiting for, one Sydney built AI agent platform listed on Microsoft's global marketplace, one federal AI policy clock starting on Monday for every APS agency including the Department of Education, and two pieces from the academic press on what universities should actually do about the technology the rest of the week is procuring.
NAPLAN 2026 school summary reports land at every school on Monday
The Final School Summary Report Results landed in every Australian school's tenant on the ESA managed assessment platform on Monday 15 June, the first set carrying the writing data alongside reading, numeracy, spelling, and grammar. The writing scripts opened on the same school side dashboards, with state and territory test administration authorities running the equivalent process in their own jurisdictions. The release covers about 4.5 million online tests taken by 1.3 million students across more than 9,300 campuses and schools in March 2026 (per the ACARA 27 March 2026 testing concludes media release). Individual Student Reports go to parents and carers through schools early in term 3, and ACARA publishes the National Results on the My School website in early August. The ACARA NAPLAN reporting key dates page on the mid-June final SSSR release to schools.
Why this matters: For independent and Catholic school heads briefing the term 3 PD frame against the school's writing data, the operational read is that the campus's writing comparison set is now sitting inside the SSSR tenant against the 2025 and 2024 cohorts, and the question the deputy is asked at next week's executive meeting is what the data layer the school's Sentral, Compass, or Schoolbox install has actually written back into the school's own performance database. For state Department of Education CIOs running the Sentral, Compass, and Schoolbox integrations behind every government school's reporting stack, Monday's release is the operational data point the federation has spent the year staging the integration for, and the schools whose dashboard renders today are the schools whose 2026 IT roadmap held the SIS to LMS pipe together against the cybersecurity and procurement push of the last six months. For ed-tech vendors selling learning analytics products into Australian school market (Microsoft's Education Insights, Google's Education Plus dashboards, the AERO-aligned platforms), the buyer is now looking at the actual 2026 SSSR data inside the school's own tenant, and the pitch that lands in July is the one that reads the school's writing data against the rest of the school's operational data rather than asking the school to hand over the assessment file in a separate platform.
Cogniti goes global on Microsoft Marketplace with the formal listing
Microsoft's Source Asia confirmed on Tuesday 16 June that the University of Sydney's Cogniti AI agent platform, built by Professor Danny Liu's team on Azure OpenAI Service, is now live on Microsoft Marketplace as a global software-as-a-service product. The University of Sydney's parallel announcement on Monday 15 June carried the customer side detail: Leiden University in the Netherlands has rolled Cogniti to 2,000 students across four faculties with three more faculties to follow, three New Zealand polytechnics (Unitec, Manukau Institute of Technology, Toi Ohomai Institute of Technology) and Auckland University of Technology are running it with their nursing programs, and Unitec alone logged 820 student conversations in second semester 2025. Liu's framing in the Source Asia piece: "We're giving them a tool where they can control how AI can be used effectively, because they're the experts" in curriculum and teaching practice. Microsoft Source Asia on the Cogniti global launch, 16 June, and the University of Sydney announcement, 15 June.
Why this matters: For Australian VCs and DVCs (Education) reading Cogniti's Leiden adoption numbers against the institution's own 2027 AI procurement schedule, the operational read is that the platform the marketplace contract can already buy is the one a comparable European research intensive has rolled to 2,000 students in one go, and the question the next Academic Board meeting answers is whether the institution's own Azure tenant is set up to publish faculty built agents on the same procurement path. For state Department of Education CIOs running Microsoft 365 Education across thousands of school tenants, Cogniti's Marketplace publication is the procurement read that the school side AI agent question (Corella, NSWEduChat, ClassmAIte, SA EdChat) now sits next to a research intensive university built product the same tenant can install, and the term 3 conversation with school heads is about whether the school wants to build its own subject agent against the same provisioning the university sector has already validated. For ed-tech vendors selling AI products into the Australian university and school sector, the Cogniti marketplace listing is the competitive read that the local research intensive can publish through Microsoft's distribution against the same procurement contract the commercial product is asking the buyer to sign separately.
DTA's AI policy mandate starts on Monday for every APS agency including the Department of Education
The Digital Transformation Agency's updated Policy for the responsible use of AI in government version 2.0 hit its six month mark on Monday 15 June, on the way to the full AI impact assessment requirement that takes effect 15 December for in-scope use cases at every non-corporate Commonwealth entity. The Department of Education is explicitly inside the policy perimeter, and the policy requires a strategic approach to AI adoption, an operationalised responsible-use process, designated accountability for AI use cases, and risk-based use case actions, with the new AI impact assessment tool now published as the contractual document the December deadline is read against. The DTA also issued updated AI procurement guidance under the same v2.0 release. Regulations.AI on the Australian DTA AI policy update, impact assessment tool, and procurement guidance.
Why this matters: For the Department of Education CIO running the 10 year digital investment strategy the iTnews August 2025 piece named (the same $51 billion annual education payments system), the operational read is that Monday's requirement is the gate the department's own AI projects (the federal Copilot trial, the up to $500,000 AI allocation last financial year) now have to be filed against, and the December assessment tool deadline is the audit trail the next Senate Estimates question will be asked against. For state Department of Education procurement directors reading the federal policy as the de facto template the next state policy will land against, the Monday milestone is the artefact the state's own AI procurement governance has to reflect by the start of term 4, and the schools that have already chosen a state-built AI tool (NSWEduChat, SA EdChat, Corella, ClassmAIte) are the ones whose use case impact assessment paper is already half written. For ed-tech vendors selling AI products into Australian Commonwealth and state agencies (the Microsoft, Google, OpenAI, Anthropic Education plans, the Australian built platforms), the December impact assessment tool is the buyer's first contractual evaluation document, and the vendor whose product cannot be filed cleanly inside the DTA's framework is the one the next procurement cycle pushes out of the conversation.
UNE academics tell the HEDx podcast universities should lead their own AI disruption
The HEDx podcast at Campus Review on Monday 15 June ran University of New England's Associate Professor Aaron Driver (head of the institution's applied AI lab) and Executive Director Future Students, Engagement and Technologies Neil Durant on the argument that universities should "lead the disruption themselves" rather than be disrupted by external AI vendors. The substance of Driver and Durant's pitch (per Martin Betts' framing): the institution that waits to be told what to do by the next platform release is the institution whose 2028 budget answers to the vendor's roadmap rather than its own pedagogy, and the institution that builds the in-house AI lab on regional Australia's actual student need is the institution whose research agenda holds against the metropolitan sandstones. Martin Betts' HEDx piece in Campus Review, 15 June.
Why this matters: For Pro VCs (Education) writing the 2027 institutional AI strategy against the ATEC implementation paper that closes consultation Monday 22 June, the UNE position is the operational alternative to the rapid procurement frame the consultation paper assumes, and the institution that holds Driver's framing is the institution whose research office is briefing the committee that the in-house lab is the cheaper procurement option for the eight year horizon. For regional university VCs (Charles Sturt, Federation, Southern Cross, USC, James Cook) reading the Stephen Matchett harbourside divide piece in Future Campus on 14 June against the institution's own low SES enrolment number (around 30 per cent at Charles Sturt and UNE, against 7 per cent at the harbourside Big Four), the AI lead the UNE applied lab is taking is the case file the institution has been waiting for to make the public argument that the regional intensive is the AI native institution the metro brand cannot become without losing its existing student profile. For ed-tech vendors selling AI products into Australian universities, the UNE in-house lab argument is the procurement read that the next contract conversation includes the build-versus-buy line the previous five years did not, and the vendor whose product cannot be configured against the institution's own applied lab roadmap is the one the next purchase order does not name.
AARE researchers say ChatGPT is the wrong tool for Indigenous education methods
EduResearch Matters ran a piece by Mitchell Rom, Tamika Worrell, Christopher Duncan, Lily-claire Deenmamode, and Ren Perkins (researchers across Australian Catholic University, Macquarie University, and the University of Queensland) on Monday 9 June arguing that pre-service teachers' use of generic generative AI tools to build Indigenous education resources undermines the relationality the Australian Professional Standards for Teachers 1.4 and 2.4 require. The substance of the argument: ChatGPT generates culturally inappropriate materials that commodify and oversimplify Indigenous knowledge, and the relationality the standards require cannot be outsourced to a tool that has no accountability to community or Country. The authors propose AI be used as a research aid (locating pre-vetted resources catalogued by AIATSIS in its 2022 Guide), with the resource selection decision retained by the human teacher. The capstone, quoting cultural lawyer Terri Janke: "AI got no dreaming." The EduResearch Matters piece on ethical AI in Indigenous education at university, 9 June.
Why this matters: For Deans of Education writing the 2027 ITE program against the AITSL Standards 1.4 and 2.4, the Rom team's argument is the case file the next program accreditation has to commit to, and the institution that publishes the position is the institution whose graduates are the ones the system trusts to teach Indigenous content. For state Departments of Education procurement directors reading the AITSL standard against the next round of approved generative AI tools (the curated lists the NSW, Victorian, and SA departments publish), the operational read is that the tool selection criterion the next list has to add is the one Rom names: the resource the AI generates is read against the AIATSIS catalogue rather than served up as authoritative, and the procurement audit trail names the human in the loop on Indigenous content as the standards trigger. For ed-tech vendors selling AI tutor and lesson planning products into Australian schools and ITE programs (the MagicSchool, Eduaide, Khanmigo, and the local equivalents), the read is that the buyer's evaluation now includes a cultural responsiveness column the product side has not been asked to answer before, and the vendor whose product cannot show the AIATSIS aware sourcing trail is the one the next state procurement form does not approve.
The thread the Wednesday read pulls together is that the system has its operational data, its formal AI procurement vehicle, its mandatory policy frame, and two arguments from the academic press about what to do next, and the answer that holds against all four is the one that names the institution's existing data, its existing teaching practice, and its existing relationships as the thing the AI is supposed to be in service of rather than the other way around. More on the operational read at digitalattitudes.com.au.