AI literacy is the bridge between fear and the graduates we need

Debating whether students should use artificial intelligence now feels like debating whether they should use the internet. Students have moved faster than institutions, with recent surveys indicating nearly 80 per cent of students in Australia using GenAI in some form, and one UK survey finding that 94 per cent use it to support assessed work. 

These numbers won’t surprise anyone in the higher education space. And they are the mandate for universities to move from AI prohibition to literacy, assurance and equitable access, while acknowledging the risks and uncertainties, and that we’re operating in an environment changing faster than traditional education models can adapt.

• Why AI literacy must be discipline specific
• Why your AI training programme won’t create an AI-literate university
• Reaching for the pizza? How to help students make informed choices with GenAI

Universities can help their communities meet this seminal moment with curiosity rather than fear. Like all learning, AI literacy is the bridge.

In my role as Australia’s first pro vice-chancellor for artificial intelligence, I see AI literacy as a core graduate capability, not a technical add-on. It means understanding what AI can and cannot do; asking better questions; checking outputs against evidence and disciplinary standards; recognising data, privacy and intellectual property risks; disclosing use honestly and remaining accountable for final judgement. It means knowing how to use AI – and when not to use it. 

A personal trainer for the mind

Used poorly, AI encourages cognitive offloading: students outsource the thinking and mistake a fluent answer for learning. In that mode, the tools become a forklift for cognition, carrying the intellectual weight for them. Used well, it acts like a personal trainer for the mind: helping students attempt harder problems, receive faster feedback and build capability through practice. The task for universities is to design learning so AI does not remove cognitive effort but redirects and deepens it.

We have to change assessment accordingly. If assessments reward only polished text, we should not be surprised when students use machines that produce polished text. To encourage cognitive amplification, educators need to assess the thinking behind the artefact: the brief students set, the contextual data used, the assumptions they challenged, the outputs they rejected, the ethical risks they identified and the reasoning behind the final answer. That might mean AI use statements, prompt trails and reflections. It could also mean oral defences, live demonstrations, practical tasks, work-integrated projects or in-class checkpoints that assure the human learning behind the assisted output.

In medicine, law, engineering, education, business and the creative industries, graduates will increasingly work with AI systems. They must learn to supervise those systems, not be supervised by them, while carrying professional responsibility for accuracy, safety and consequence.

At my university, this principle sits behind our AI-first strategy. Through our landmark partnerships, we are providing access to frontier AI tools to everyone at scale, beginning with 5,000 licences in 2026 and extending free access to advanced AI tools for all 40,000 students and staff by 2027. We will work with industry partners to embed AI into the curriculum with the goal to give every student, regardless of background, discipline or location, access to the tools and teaching they will need to compete and build AI literacy by applying it repeatedly across their respective courses. 

Accessible AI

Access at scale matters, because otherwise, AI literacy risks widening existing inequities. Some students will arrive having used advanced tools for years. Others will have been warned away from them, priced out of them or unsure about what is permitted. Without a safe, common foundation, universities risk creating two groups of graduates: those who have practised working with AI under expert guidance, and those who have only experimented privately and anxiously.

The same logic applies to staff. Academic and professional staff cannot model good AI use if they are left to build capability alone. They need safe tools, practical playbooks and communities of practice. In a distributed institution, no central team can “own” AI literacy on behalf of everyone. 

The right operating model is a coordination layer: setting a common framework, supporting local AI use, celebrating AI champions across the university, cross-pollinating ideas and helping experiments become scalable practice. We’ll see the proof points locally: in redesigned feedback, AI-enabled industry projects and better student support. 

Employers are already signalling the need for AI-fluent talent. PwC’s 2025 Global AI Jobs Barometer found that workers with AI skills command a 56 per cent wage premium on average. This shows that AI is not confined to technology courses but belongs in every degree, because AI is changing every field. 

The graduates who stand out in the future jobs market will be those who were taught how to use AI to think better, learn deeper, work faster, act ethically and create more value. Our responsibility is to make that capability teachable, assessable and visible before they leave us.

Phil Laufenberg is pro vice-chancellor (artificial intelligence) and chief artificial intelligence officer at La Trobe University.

If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.