Google Gemini defaults on in Classroom from 10 August

Digital Attitudes, 6 August 2026

Thursday of Term 3 Week 4, and three of this week's biggest moves land on the same tenant admin: Google switches Gemini in Classroom on by default for every student on Monday, OpenAI drops a set of role based education plugins into ChatGPT Edu, and the EU AI Act's chatbot disclosure obligation goes live for any vendor with a European footprint.

Google switches Gemini on for every Classroom student from 10 August, opt out is per organisational unit

On 4 August, Google Workspace Updates announced that Gemini in Classroom will expand to K-12 and higher education students of all ages who have been granted access by their administrator. The web rollout starts on Monday 10 August, mobile follows on 17 August, and the setting is on by default for teachers and students on Fundamentals, Standard, and Plus education editions. The new build ships contextualised starter prompts (Gemini pulls in the specific class, assignment title, instructions, and curriculum materials as context), a "Learn with Gemini" hover option on the redesigned Due Soon homepage, and a one click sync into Gemini Notebook so a teacher's materials become interactive study guides, audio overviews, and infographics. Administrators who want under 18s excluded must build a separate organisational unit for those students and turn Gemini in Classroom off for that OU alone (Google Workspace Updates on the Gemini in Classroom expansion).

Why this matters: For every Australian school running Google Workspace for Education, this is a Monday morning tenant decision, not a term 4 planning item. The default on posture flips the burden of proof onto the ICT lead and the principal: if the school wants to keep primary or junior secondary students out of Gemini in Classroom, the organisational unit for under 18s has to exist and the toggle has to be off before web rollout hits on 10 August. The concrete Term 3 action for a director of digital learning is a one page brief for the SLT this week that names three things, whether the school's Google OU structure separates under 18s cleanly, whether the current AI policy actually permits a default on chatbot inside a graded assignment, and whether the Department of Education guidance the school aligns to (VIC opt in, NSW EduChat only, QLD Corella positioning) is compatible with the Google default. For a Catholic system or independent group ICT lead, the same brief needs to name whether a system wide instruction to disable at the tenant admin level is going out to member schools before Friday, because "we assumed schools would opt out per class" is the position that will not survive the first parent question in Week 5.

OpenAI drops three role based plugins into ChatGPT Edu and ChatGPT for Teachers

On 4 August, OpenAI added K-12 Educator, College Educator, and College Student plugins to ChatGPT Edu and the free ChatGPT for Teachers deployment. The K-12 Educator plugin handles lesson planning, differentiated resources, interactive visuals, and workflow items like assignment translation, exit ticket briefs, family updates, and practice tests. The College Educator plugin covers syllabus updates, interactive website creation, multimedia assessments, and packaging content back into the LMS, and connects to calendars, documents, and approved institutional tools. The College Student plugin, grounded by OpenAI in learning science principles, turns course materials into flashcards, quizzes, guided tutoring, and visual explanations, and institutions keep control over tool permissions and access inside managed workspaces (EdTech Innovation Hub on the OpenAI education plugins launch).

Why this matters: Tuesday's brief flagged the 10,000 seat ChatGPT for Science rollout at nine Australian universities. This is the operational layer that sits on top of it, and it also lands on any AU K-12 school running ChatGPT for Teachers. For a DVC (Academic) or a director of digital learning, the College Educator plugin is the one that changes the question from "should staff use ChatGPT" to "should staff push course materials into an assistant that repackages them for the LMS", which is a Canvas or Blackboard data flow question the current AI policy probably does not answer. The concrete Term 3 action for a DVC (Academic) is a one page note to the academic board on whether the College Student plugin's flashcard and tutoring outputs count as "AI assisted study aids" (allowed) or "AI generated assessment prep" (contested), because that call is a Semester 2 academic integrity policy edit, not a Q4 one. For a K-12 curriculum leader on ChatGPT for Teachers, the K-12 Educator plugin's Learning Commons alignment is US standards not Australian Curriculum v9.0, and the translation step should be on the agenda before any staff PD points the whole staffroom at it.

EU AI Act Article 50 goes live on 2 August, education high risk pushed to December 2027

On 2 August, the EU AI Act's Article 50 transparency obligations became enforceable, and the Digital Omnibus (in force from 27 July) pushed the Annex III high risk deadlines for education AI (student assessment, admissions screening, exam and behaviour monitoring) back to 2 December 2027. Article 50(1) requires that any system designed to interact directly with people must inform users they are interacting with an AI system, "unless this is obvious from the circumstances", and the compliance analyses read the obvious clause narrowly (a natural language chatbot without an explicit disclosure likely fails). Article 50(2) requires machine readable marking of AI generated outputs with a grace period to 2 December 2026, and Article 50(4) requires deployer disclosure of deepfakes and public interest text unless the content underwent documented human editorial review. Penalties for GPAI provider breaches reach the higher of 15 million euro or 3 per cent of worldwide annual turnover (Compliance Hub on what actually applies from 2 August under the AI Act and Omnibus, Gibson Dunn on the Omnibus agreement postponing high risk deadlines).

Why this matters: The AU sector reads this two ways. First, any Australian university with EU based cohorts, EU campuses, or EU headquartered ed-tech vendors picks up the Article 50 chatbot disclosure obligation immediately, which is the compliance line the OAIC's own ADM guidance (due September, in force 10 December) will be measured against. Second, the Annex III postponement to December 2027 gives the assessment and proctoring vendor stack 16 more months to land conformity assessments and human oversight documentation, so a DVC (Academic) buying an AI proctoring product in Term 4 should ask the vendor to name which of Articles 8 to 27 the product is already engineered for and which are still on the roadmap. The concrete Term 3 action for a university general counsel is a one page check on which of the tenant's student facing tools (LMS chatbot, library assistant, careers bot, IT service desk bot) currently ship without an explicit "you are talking to an AI" disclosure, because the same disclosure line will land in the AU ADM policy update in December and rewriting all the tool copy twice is the avoidable spend.

Three tenant admin decisions in one week, and the one that lands first (Google's default on switch on Monday) is the one most schools are still treating as a Q4 policy question. More at digitalattitudes.com.au.