Deakin University

First, I would want to see clear, practical commitments that move beyond broad agreement. That means shared baseline expectations for safe, trustworthy AI and concrete steps for transparency, accountability, and human oversight that can actually be implemented across sectors. Second, success would include meaningful alignment across governance approaches. Not full uniformity, but enough interoperability that organisations, educators, and developers are not navigating completely fragmented systems. Without that, progress becomes slow, inconsistent, and harder to scale responsibly.

• Safe, secure and trustworthy AI
• Transparency, accountability, and human oversight
• Interoperability of governance approaches

As someone teaching AI, I see a disconnect that's becoming harder to ignore. Most courses teach students how to build models, optimise them, and deploy them. But far fewer teach them how to question what they're building, who it affects, or where the boundaries should be. That's why safe, secure and trustworthy AI matters so much to me. It's not just about technical performance. It's about whether students understand the responsibility that comes with creating these systems. The same applies to transparency, accountability, and human oversight. I want my students to ask harder questions. Why did the model make this decision? Who is accountable when it fails? Where does human judgment need to step in? These aren't side topics. They are core skills. Interoperability of governance approaches also matters because my students won't work in a single system or country. AI doesn't respect borders, and neither do its risks. They need to understand how different governance frameworks connect and where they clash. Right now, capability is accelerating, but governance awareness is not keeping pace. I don't want to graduate students who can just build powerful systems. I want them to build responsibly, with awareness, judgment, and intent.

What I see most clearly is a growing gap between capability and responsibility in AI education. Students are getting very good at building things. But they are not always being taught how to think about the consequences of what they build. Governance is often treated as an add-on, something separate from the "real" technical work, rather than something embedded throughout. There is also a gap between theory and practice. Students might learn about ethics in abstract terms, but they are rarely asked to apply it. They are not consistently trained to assess risk, document decisions, audit systems, or take accountability in realistic scenarios. Another challenge is that many educators are still figuring this out themselves. The technology is moving fast, and governance frameworks are still evolving. That makes it harder to design courses that meaningfully integrate both. If we continue on this path, we risk creating a generation of AI practitioners who are highly capable but underprepared to deal with the broader impact of their work. The question I keep coming back to is simple. Are we preparing students to build AI, or to build it responsibly?

Coming from Mauritius and now working in Australia, I see two very different stages of AI readiness, and the governance gaps are visible in both, but in different ways. In Mauritius, AI adoption is still emerging, but governance frameworks are not keeping pace even with this early growth. There is limited national coordination around safe, secure and trustworthy AI, and very little emphasis on transparency, accountability, or human oversight in practice. The risk is that AI will be adopted in fragmented ways, without clear standards, safeguards, or local capacity to evaluate its impact. Education systems are also not yet fully equipped to prepare students for responsible AI use, which could widen the gap between technology adoption and understanding. In contrast, Australia is more advanced in both AI adoption and governance discussions. However, even here, I see a gap within education. As an educator, I notice that students are becoming increasingly capable of building and deploying AI systems, but governance awareness is still not consistently embedded in curricula. Topics like accountability, risk, and oversight are often treated as secondary rather than core.Coming from Mauritius and now working in Australia, I see two very different stages of AI readiness, and the governance gaps are visible in both, but in different ways. In Mauritius, AI adoption is still emerging, but governance frameworks are not keeping pace even with this early growth. There is limited national coordination around safe, secure and trustworthy AI, and very little emphasis on transparency, accountability, or human oversight in practice. The risk is that AI will be adopted in fragmented ways, without clear standards, safeguards, or local capacity to evaluate its impact. Education systems are also not yet fully equipped to prepare students for responsible AI use, which could widen the gap between technology adoption and understanding. In contrast, Australia is more advanced in both AI adoption and governance discussions. However, even here, I see a gap within education. As an educator, I notice that students are becoming increasingly capable of building and deploying AI systems, but governance awareness is still not consistently embedded in curricula. Topics like accountability, risk, and oversight are often treated as secondary rather than core.

Right now, AI governance is developing unevenly. Some countries are advancing quickly, while others, like Mauritius, are still building foundational awareness and capacity. The Dialogue can help bridge this gap by creating shared understanding and accessible entry points for countries at different stages of readiness. One of its most valuable roles would be to support interoperability. Not by forcing a single global framework, but by aligning core principles and expectations so that systems, standards, and approaches can work across borders. This is critical because AI systems are not confined to one country, and fragmented governance makes it harder to manage risks effectively. It can also elevate voices that are often missing, especially from smaller or developing nations, as well as educators and practitioners on the ground. These perspectives are essential if governance is to be inclusive and realistic, rather than driven only by the most advanced economies.

There are already several important initiatives shaping AI governance globally, including the OECD AI Principles, UNESCO's Recommendation on the Ethics of AI, and regional efforts such as the EU's AI Act. There are also growing partnerships between governments, industry, and academia focused on responsible AI development and standards. These initiatives provide a solid foundation, but they often operate in parallel rather than in a truly connected way. From where I stand, especially coming from Mauritius and working in Australia, this fragmentation makes it harder for smaller or less-resourced countries to engage meaningfully or keep up. The AI Dialogue can add real value by acting as a bridge between these efforts. Instead of creating another standalone framework, it can help align and translate existing principles into something more coherent and accessible across different contexts. This is particularly important for countries that are still building their AI capacity and governance structures. Another gap is the connection between high-level policy and what happens in practice. Many existing initiatives outline what should be done, but not always how to implement it in education systems, organisations, or everyday AI development. The Dialogue could help turn principles into practical guidance, especially for sectors like education where governance is still not fully embedded. It can also bring in voices that are often underrepresented, including educators, students, and countries in earlier stages of AI readiness. That inclusivity would strengthen global cooperation and make governance frameworks more grounded.