In Australian universities, AI is not a skill that is acquired only after graduation. A survey commissioned by OpenAI found that 81% of the surveyed students use AI multiple times per week, 33% use it several times a day, and 28% have even created intelligent agents to assist with their studies, work, or daily life. With students bringing these tools onto campus, the issues faced by schools have changed: no longer is the question merely whether to allow it or not, but how to ensure that widespread use does not undermine independent judgment and professional training.
On November 5th, OpenAI announced its collaborations with the University of New South Wales, the University of Sydney, La Trobe University, and the University of Adelaide. The most significant collaboration comes from the University of New South Wales, where ChatGPT Edu will cover over 80,000 students and staff members. The University of Sydney plans to make it available to all students in the first semester of 2027 after a pilot period; La Trobe University already has 3,500 employees authorized to use it and plans to cover 45,000 students and staff in the first quarter of 2027.
Account popularization is just the outermost layer.
If we only look at the number of accounts, it seems like a routine campus software procurement. However, these schools are integrating AI into specific training scenarios. La Trobe University’s emergency medicine course uses ChatGPT to generate clinical situations, allowing students to interpret electrocardiograms, identify cardiac events, and manage triage for mass casualties. The Faculty of Education, on the other hand, employs social robots equipped with large models to simulate conversations that may arise between future teachers and parents. Students are not just required to produce well-written essays; they must also explain their reasoning and take responsibility for their decisions.
This design reveals the true costs of scaled-up adoption. The tool can quickly generate scenarios, but it cannot define what constitutes a qualified performance for schools; it can help students practice repeatedly, but it cannot replace teachers in deciding which abilities must be demonstrated without the assistance of AI. La Trobe University has disclosed that expanding its use also comes with costs related to permission management, training, and token. A 10% monthly increase in active users means that back-end governance must keep up with the enthusiasm on the front end; otherwise, "available to everyone" can easily become a situation where only a few people are proficient in using it, while others are left with just a login entry point.
The University of Sydney has extended this same approach to infrastructure. In addition to rolling out ChatGPT Edu in phases, the university has also established a $1 million AI Sustainability Research and Innovation Fund to observe the impacts on energy, cooling systems, and the product’s lifecycle. There is a change that is often overlooked: when AI shifts from being a personal subscription to becoming a campus public facility, its cost is no longer just a software bill; it also includes data governance, environmental monitoring, teacher training, and course rewrites.
Starting from this exam, there will be a distinction between "knowing how to do something" and "being able to apply that knowledge in practice".
More critical actions occur in the evaluation rules. The University of New South Wales has introduced the Can, Can, and Must frameworks, allowing teachers to clearly specify what AI can be used, what cannot be used, and what must be used according to learning objectives. The University of Sydney, on the other hand, separates the verification of independent abilities in a supervised environment from the verification of collaborative abilities in open assessments. The rules do not completely exclude AI, nor do they equate the use of tools with mastery of knowledge.
This division allows for both issues to be examined simultaneously. Can a student demonstrate their foundational knowledge without relying on model answers? And when the use of AI is permitted, can they ask appropriate questions, verify the results, and accept the consequences? The former safeguards the core of academic learning, while the latter prepares students for the real-world work they will face in the workplace. If only one of these aspects is emphasized, schools will either produce individuals who are adept at taking exams but unable to utilize new tools, or those who can generate answers but lack the ability to assess their correctness.
The survey sample consisted of 1,018 Australian university students aged 18 and above, with the interviews conducted from August 3 to 7. It was commissioned by OpenAI and cannot be considered an independent census of the entire higher education system, but it is sufficient to illustrate why the participating schools included both "use" and "judgment" in their plans. 79% of the respondents hoped that schools would provide more opportunities for them to use advanced AI tools to solve real-world projects; the same proportion of people wanted to use AI for health, climate, science, or cybersecurity issues.
After the university expansion ChatGPT, the biggest change was not the increase in the number of accounts in the computer labs, but rather the structure for demonstrating competence. Students need to provide evidence in both environments with and without tools, and teachers need to translate course objectives into clear boundaries for usage. AI has transformed from an extracurricular shortcut to a formal training facility, and as a result, the school's approach has shifted from "preventing students from using" to "designing how to use it effectively as a form of learning."
