ChatGPT for Teens defaults on in Australia Tuesday

Digital Attitudes, 6 September 2026

Sunday, and the shape of the week ahead is three defaults arriving on top of each other. A vendor default that flips on Tuesday, a NESA operational deadline that starts the same day for NSW schools, and a benchmark that measures how ready the sector actually is. The three stories together read as a single question for a Term 4 planning meeting: is the policy stack keeping up with the vendor stack, or not.

ChatGPT for Teens flips on by default for Australian teens from Tuesday, and parents get notifications but not transcripts

OpenAI announced ChatGPT for Teens on Tuesday 18 August 2026 with a global rollout it expected to complete within two weeks, and Security Brief Australia confirmed the Australian availability from Tuesday 8 September 2026 (Security Brief Australia on the Australian rollout, OpenAI on ChatGPT for Teens). Any Free or paid personal ChatGPT account where OpenAI estimates the user is under 18, or where the user states an age between 13 and 17, is automatically placed into the Teen experience. That experience defaults to Study Mode style guidance, Quiet Hours (a scheduled period where a teen should not use ChatGPT), tighter content limits (self harm, violence, eating disorders, dangerous activities, explicit content), an under 18 model specification that rules out romantic language and discourages emotional dependence, and break reminders. Parents with a linked account can set Quiet Hours, turn Study Mode on directly from the parental controls, and receive safety notifications in limited high risk situations. Parents cannot read or monitor the teen's conversations. Brent Thomas, Head of Policy for Australia and New Zealand at OpenAI, on the record: "We want teens to use AI responsibly to learn, create and explore".

Why this matters: for a Term 4 Acceptable Use Policy, the question has shifted from "is ChatGPT allowed for students" to "which version of ChatGPT is the student actually seeing, and does the school know". Two operational reads. First, the audit question. If an under 18 student signs into ChatGPT on a school issued device on Tuesday, they enter the Teen experience whether the school has updated its AUP or not. The safer default is genuinely safer than the adult one, and that alone changes what an integrity officer can plausibly claim about student exposure. It also changes what a school can plausibly say to parents, because the parental controls that ship with the change give notifications but not transcripts, so a parent linking their child's account does not become a monitoring proxy for the classroom. Second, the record keeping read. NAPLAN 2026's equity gap analysis (yesterday's brief) rested on disaggregated data. The equivalent for AI in the school is a live count of which students are using which tools under which account. If the school data warehouse cannot answer "how many under 18 students in our tenant are logged into a personal ChatGPT account", the Term 4 AUP will be an aspiration, not a control.

NESA's take-home cap hits NSW Stage 6 assessment plans this week, and school management vendors have three weeks to catch up

The NESA take-home changes announced by NSW Acting Premier Prue Car on Monday 1 September 2026 begin biting into HSC school-based assessment programs from Term 4 2026, with the Preliminary equivalent starting Term 1 2027, and NESA has committed to updating the ACE Rules and syllabus documents before the end of Term 3, which is three weeks away (NSW Government on the new rules for AI in take-home HSC assessments, Education Matters Magazine on the new take-home rules). The rule limits HSC and Preliminary school-based assessment programs to a maximum of one take-home task at a maximum weighting of 15 per cent, or 7.5 per cent of the total HSC mark. Courses with a major project or an HSC practical exam are exempt. Schools still decide the number, type and weighting of all in-school tasks.

Why this matters: for a NSW deputy principal building a 2027 assessment plan, and for the head of teaching and learning validating the Term 4 program in Compass, Schoolbox, Sentral or SEQTA, this is the operational trigger point. Two reads. First, the data audit. Any assessment task recorded in the school management system as a take-home in a Stage 6 course now has to be counted per course, checked against the 15 per cent cap, and either kept, reweighted, or replaced with an in school equivalent. If the system does not carry take-home status as a queryable field, the audit is a manual read of every assessment notification for every Stage 6 course, and that is a two day job the deputy principal does not have on the way into Term 4. Second, the vendor conversation. School management vendors have been quiet on assessment classification. If a school runs the audit and finds the platform cannot filter for take-home tasks, that is a product gap worth raising formally, because Preliminary limits land Term 1 2027 and the change will not go away. The three weeks between now and the Term 3 syllabus updates is the window to run the audit, log the vendor requests, and rewrite the affected task notifications before Term 4 opens.

Two thirds of Australian education providers still have no documented AI policy, ReadyTech benchmark finds

The ReadyTech AI and Innovation Benchmark 2026, surveying 96 Australian education organisations across schools, VET and higher education, reports that 94.8 per cent are either exploring or already using AI, 28.1 per cent use it regularly in at least one core workflow, and more than 40 per cent have deployed generative AI organisation wide, yet only 34.4 per cent have a documented AI use policy and 30.2 per cent have a responsible AI framework (The Educator K/12 on the ReadyTech benchmark, ReadyTech on the AI and Innovation Benchmark 2026). 11.5 per cent said they have no governance measures in place. On academic integrity, 43.8 per cent expect the problem to worsen in the next two years, while 77.1 per cent report they are moderately or very prepared to respond. 42.7 per cent have deployed AI detection tools. 32.3 per cent run student AI literacy programs. 70.8 per cent plan to increase AI spending in the next 12 months, and 67.7 per cent see AI as more opportunity than threat.

Why this matters: the number that reads oddest is the pair of 34.4 per cent (has a documented policy) and 77.1 per cent (prepared to respond). A leadership team that rates itself prepared without a documented policy is running on confidence, not on written control. Two operational reads for a school business manager or a university pro vice chancellor academic. First, the board paper question. If a board asks in October "do we have an AI policy" and the answer is "no, but we are prepared", the follow up is "prepared how, by whom, and against what" and the room goes quiet. The Term 4 defence is a one page written policy that names which tools are approved, where student data can and cannot go (nothing sensitive into an unapproved model, per NSW and Victorian departmental guidance), and what a teacher or lecturer does when they suspect a submission is AI generated. Anything longer than a page is unlikely to be read, so it will not be followed. Second, the detection tool line. 42.7 per cent have deployed AI detection, at a moment when Curtin University disabled Turnitin's AI detection from 1 January 2026 and the University of Queensland disabled it in mid 2025, both citing reliability. Deploying detection is not the same as trusting it, and a school or university buying an AI detection tool in the next 12 months needs to answer for how a false positive will be handled before the first one lands on a student's file.

Three defaults, one Sunday. A vendor default flips on Tuesday. A NESA operational deadline starts the same day for NSW schools. And a national benchmark says most providers are running on confidence rather than a written policy. The Term 4 planning meeting worth having this week is the one that closes the gap between the three. Prior editions and the fuller operational read live at digitalattitudes.com.au.