Universities are racing to embed AI literacy from first year. This is a response to the reality that nearly 80% of Australian students already use generative AI in their studies, often without institutional guidance, and employers want graduates with at least some level of AI literacy. On paper, starting early and scaffolding skills across a degree makes sense. However, in practice, it is not that simple.
As an education designer and a member of our AI working group, I have been involved in building frameworks, resources and workshops to bring AI literacy into first-year subjects. The challenges are both practical and conceptual. Students often treat AI activities as extra work when they feel disconnected from the subject. Some see little value in comparing their writing with AI outputs or reflecting on the tool’s role. Others raise ethical concerns. Educators question whether AI fits their discipline or worry it undermines core skills. When AI is not clearly tied to subject learning outcomes or assessment, and simply adds to workload, its value is hard to see.
But there is a deeper problem. A central message of AI literacy is that students must not simply trust what generative AI produces. They are expected to evaluate outputs, interrogate the data, recognise limitations and bias, question suggestions, and know when not to rely on the tool. All of this requires knowledge and skills.
A first-year student still building foundational understanding may not recognise an incomplete or misleading explanation, spot weak or biased claims, or know which parts of an AI response need checking. To work critically with AI, students need to grasp the relevant concepts, apply domain expertise, and make informed judgements about whether, and how far, to trust what the tool produces. These are disciplinary thinking skills, not AI skills.
This creates a paradox: We ask first-year students to critically evaluate AI precisely when they are still developing the knowledge and judgement that make evaluation possible.
Part of the difficulty is that two different ideas are being collapsed: learning about AI and learning with AI.
Learning about AI means understanding what generative AI is, what it can and cannot do, its limitations and risks, and how to use it ethically. There is a strong case for introducing this early, especially when students are already using the tools.
Learning with AI is different. It means using AI to perform disciplinary tasks, evaluating what it produces, and deciding when to accept, question or reject its suggestions. That requires knowledge and skills of the discipline itself.
Many universities are trying to do both from day one. For first-year students this is often unrealistic. Learning about AI, guidance on responsible use and open conversations with educators, can start early. Using AI as a learning tool before students can judge its outputs makes the task harder than it needs to be.
Some institutions have already drawn a similar line. In July 2026 the University of Chicago Law School barred phones and laptops in first-year core classes and required students in legal writing to produce initial drafts without AI before tools are introduced later. The following month, UChicago’s Social Sciences Division moved to broader ‘analog’ restrictions in core courses so students learn to read, write and think independently. Outside the US, the University of the Philippines College of Law is restricting AI and devices in first-year classes from 2026-27, citing concerns about declining writing, comprehension and independent analysis.
More recently, UC Berkeley Law has adopted a strict default policy that applies across all years of the degree. It prohibits AI use for conceptualising, outlining, drafting, revising or editing any work submitted for credit, and bans AI entirely in exams.
Critics argue that delaying AI use simply exports the learning curve to a graduate’s first employer. Even they usually accept that some foundational tasks, a first legal writing draft, a closed-book exam, should remain AI-free to protect learning.
A more staged approach is possible. In first year, the priority should be building strong disciplinary foundations while opening ongoing conversations with students about AI, its capabilities, limitations, ethical issues and responsible use. This can include short foundational modules, class discussions, and guidance tools such as the AI Assessment Scale so students understand how AI may (or may not) be used in their assessments. Full integration of AI as a learning tool across the curriculum and teaching can then begin more deliberately from second year, once students have had time to develop the knowledge and skills needed to evaluate its outputs and use it critically.
This also matches what industry is asking for. PwC Australia’s 2026 AI Jobs Barometer found that as AI takes over routine tasks, junior workers are expected to handle more complex work and use AI responsibly. Disciplinary knowledge, critical thinking and judgement become more important, not less. As PwC’s Peter Wheeler put it, the workforce of tomorrow will be “less ‘doing’ and more ‘thinking’.
The question universities should ask is not simply whether first-year students need AI literacy, but when they should begin learning with AI, and when that use genuinely contributes to their learning.
If the goal is graduates who can work with AI critically and responsibly, they need something to think with: knowledge, disciplinary understanding, and opportunities to develop their own judgement. AI literacy matters, but learning about AI and learning with AI are not the same thing. Sometimes the best preparation for working effectively with AI is the chance to learn and think without it first.
Regards,
Sara Abdelmawgoud
Leave a Reply