AI in Schools
CurricuLLM, Lindisfarne and AERO each ship their evidence
The Tuesday read is three pieces of evidence that landed in the past fortnight, each from a different corner of the schools sector: one Australian ed-tech vendor that has decided what AI safety in a K-12 classroom should actually look like, one independent school that has written up the procurement bet it made on devices, and the national research body that has put a 53-study evidence base on the table for the cohort most likely to be flagged by an AI-mediated tool in the next term.
CurricuLLM ships a 21-category student safety spec for AI in classrooms
Sydney ed-tech company CurricuLLM announced on 11 May the launch of its Student Safety Centre, a monitoring layer that sits across student-AI interactions in K-12 classrooms and identifies risks across 21 harm categories including academic offloading, emotional attachment to AI tools, and personal disclosures. The product carries Apgard YouthSafe AI certification and has been assessed against the Education Services Australia Safer Technologies 4 Schools framework. CEO Dan Hart's framing is that "generic content filters miss almost everything that actually matters in a school setting," and the design principle is "minimum necessary visibility": flagging concerning content while keeping routine learning conversations private. CurricuLLM reports 89 per cent accuracy on Australian curriculum content against 41 per cent for general-purpose AI models. IT Brief Australia on the Student Safety Centre launch.
Why this matters: This is the first Australian-built product to put the seven NSW IPC privacy tests, the December Privacy Act automated decision-making transparency requirement, and the OAIC Children's Online Privacy Code drafting work into a single line item with a price. For independent and Catholic school CIOs working through the readiness checklist before signing off on Copilot, Gemini for Education, or ChatGPT Edu inside classrooms, the procurement question is whether the existing vendor stack already covers the 21 harm categories or whether a separate safety layer is part of the buy. For the state systems, the read is similar: NSW EduChat, SA EdChat and WA ClassmAIte already have content filtering baked in, but the cross-sector buyer comparison just got cleaner. The product is also the sector-side response to the cognitive offloading argument Professor Leslie Loble and Professor Jason Lodge put back into The Educator yesterday, because "academic offloading" is the first harm category on CurricuLLM's list.
Lindisfarne writes up its BYOD to 1:1 Apple pivot, and the device cycle just got sharper
The Educator's 25 May write-up of Lindisfarne Anglican Grammar School's two-year transition from a Bring Your Own Device model to a 1:1 Apple environment is the operational case study other independent schools were waiting for. Principal Stuart Marquardt and Head of Professional Practice Dr Kane Bradford describe a staged rollout with iPad in junior years moving to M4 MacBook for senior students, a two-year device rotation cycle, and Year 12 students preparing for the HSC in a 2:1 environment (iPad alongside MacBook). The school took ten years to land on Apple as the single platform and picked up the Australian Education Awards Best Use of Technology in 2023 and an Excellence Award in 2025. The Educator on Lindisfarne's 1:1 Apple transition.
Why this matters: For independent and Catholic school business managers running the 2027 device cycle decision, the Lindisfarne write-up is the first detailed Australian case for going single-platform 1:1 rather than continuing the BYOD model that has dominated the sector since the early 2010s. The procurement read is the staged-age progression (iPad junior, hybrid mid, MacBook senior) and the two-year rotation cycle, both of which sit at the higher end of the per-student cost line but reduce the support load on the IT team and the configuration drift teachers spend the first five minutes of every lesson managing. For schools currently running mixed Windows and Apple fleets, the question to take into the executive meeting is whether single-platform standardisation produces the operational return Lindisfarne is reporting, and what the equivalent Microsoft Surface-only argument would look like in the same school. The case study is useful because it names the trade-offs, which most vendor-funded case studies do not.
AERO puts 53 studies on the table for Autistic and ADHD teaching practice
The Australian Education Research Organisation released its discussion paper Teaching practices to support Autistic students and students with ADHD on 20 May, drawing on an umbrella review of 53 studies published between 2014 and 2025, plus targeted literature searches, First Nations-focused research, expert consultation, and policy scans. The paper sits inside AERO's broader evidence-base work and is the most current cross-jurisdictional synthesis the Australian sector has on practice that works for neurodivergent learners. The AERO discussion paper on teaching practices for Autistic students and students with ADHD.
Why this matters: For school IT and curriculum leads who buy or recommend adaptive learning platforms (Mathletics, LiteracyPlanet, Reading Eggs, Education Perfect, the various PAT-aligned products) and assistive technology, the AERO paper is the next evidence document a tender response will have to engage with. The procurement read is whether the platform's adaptive logic does what the evidence says works for neurodivergent learners, or whether the "personalisation" claim is the older content-pacing kind that does not. For senior leaders watching the AERO-into-Teaching-and-Learning-Commission restructure announced earlier this month, the timing of the paper is a useful signal of what AERO research outputs will look like after the merge, and what the new Commission will be expected to maintain as a sector evidence base. The thread back to the CurricuLLM story is direct: an AI-mediated platform that flags a neurodivergent learner under "academic offloading" without engaging the AERO evidence base will be the kind of vendor decision a principal cannot defend twelve months from now.
The thread this morning is the gap between the vendor pitch and the documented operational case, in three places at once. A safety spec, a device case study, and a research synthesis all landed in the past fortnight, and the schools and systems that read all three are the ones whose procurement papers will hold up to the harder questions on the EDUtech floor next week. More on the operational read at digitalattitudes.com.au.