Independent

Success is not just about everyone agreeing, but about having a more honest and grounded conversation than we usually see in AI spaces. A good outcome would be people actually naming the tensions instead of avoiding them. Right now, AI governance conversations often stay at the level of principles. We say things like "responsible" or "ethical," but we do not always sit with the trade-offs behind those words. If this dialogue can surface where countries, companies, and communities genuinely disagree and why, that is already a step forward. It would also need to shift who is in the room and who is taken seriously. In my work across youth civic engagement, community programming, and AI policy spaces, I have seen how often the same voices dominate, while the people most impacted by these systems are treated as an afterthought. A successful dialogue would not just include those perspectives, but actually let them shape the direction of the conversation. Another important outcome is moving beyond talk. There is no shortage of frameworks or declarations. What is missing are clear pathways for accountability and coordination. Even small, concrete next steps, like shared approaches to auditing or commitments to transparency, would matter more than another set of high-level principles. Ultimately, I believe success looks like continuity. One-off conversations do not change systems. If this dialogue leads to ongoing collaboration, trust-building, and clearer roles across sectors, then it has done something real. At minimum, people should leave with a better understanding of what needs to change and what their responsibility is in making that happen.

• Safe, secure and trustworthy AI
• Transparency, accountability, and human oversight
• Protection and promotion of human rights
• Social, economic, ethical, cultural, linguistic and technical implications of AI

These four areas line up with where I see the biggest disconnect between how fast AI is moving and how little oversight actually exists. These priorities reflect a need to get more practical. Less talk about what "good AI" looks like, and more focus on what actually holds systems and institutions accountable. On safe, secure, and trustworthy AI, I'm seeing systems get rolled out before anyone has really thought through the risks. Working with youth, especially in digital spaces, it's clear that harm shows up quickly and unevenly. Safety can't be reactive. It has to be built in from the start. The broader social and ethical implications matter because AI is not neutral. A lot of my work sits across different communities and contexts, and you can see how existing inequalities just get carried into these systems. If we are not actively accounting for that, we are just scaling the same problems. Human rights is non-negotiable for me. I come into this work from a civic engagement and equity lens, so I'm always asking who is being protected and who is being left out. AI governance should not be inventing new values. It should be reinforcing existing rights and making them enforceable in digital systems. And then transparency, accountability, and human oversight is the piece that makes everything else real. Right now there is a lot of language around principles, but not a lot of clarity on who is responsible when things go wrong. From the policy work I've been part of, that gap comes up again and again.

Who actually has power in AI governance? A lot of discussions focus on risks, ethics, or technical standards, but not enough on the concentration of power across a small number of companies and countries. That shapes everything, from whose values get embedded into systems to who benefits economically. If that is not addressed, the rest of the governance conversation stays limited. Another is access and capacity, especially for communities and countries that are expected to adopt or respond to AI but are not resourced to shape it. In my work, I see this even at the local level. People are expected to engage with systems they do not fully understand and have no real influence over. Digital inclusion is often talked about as access to tools, but not as meaningful participation in decision-making. I also think implementation gaps deserve more attention. There are already plenty of frameworks, guidelines, and principles. The issue is not a lack of ideas. It is that they are not consistently applied or enforced. There needs to be more focus on what it actually takes to operationalize governance across different contexts. And then, longer-term impacts tend to get sidelined. A lot of current conversations are focused on immediate harms, which are important, but there is less space for thinking about how AI systems will reshape labour, civic participation, and social structures over time. To me, these are not separate issues. They cut across everything else and determine whether any of the existing priorities actually lead to meaningful change.

In Canada, the gap is not a lack of awareness. It is the gap between commitments and implementation. On safety and trust, AI is already being used in hiring, public services, and online platforms, but oversight is uneven and often reactive. In the youth civic engagement space I work in, this shows up as real concerns around online safety, misinformation, and emerging harms like AI-generated image abuse. The systems are moving faster than the safeguards. On social and economic impacts, Canada talks a lot about inclusion, but there is still a disconnect in practice. Communities that are already marginalized, including newcomers, racialized youth, and neurodivergent people, are more likely to experience harm or be excluded from opportunities shaped by AI. At the same time, there is a real opportunity here. Canada has strong research institutions and a policy environment that could lead on inclusive AI if it moves beyond pilot projects and into scaled action. On human rights, the challenge is that protections are not always clearly translated into the AI context. There is growing attention to this, but enforcement mechanisms are still unclear. People do not always know when their rights are being impacted, or what recourse they have. And on transparency and accountability, this is probably the biggest gap. There is still limited visibility into how many AI systems are being used, especially in public-facing contexts. From the policy work I have been part of, the issue is not just transparency, but who is responsible when things go wrong. The opportunity is that Canada is still in a position to get this right. But that depends on moving faster on governance, and grounding decisions in the realities people are already experiencing.

It can play a useful role, but only if it moves beyond being a space where everyone says the same things in slightly different ways. Right now, AI governance is happening in silos. Governments, companies, and international bodies are all working on their own tracks, and they do not always connect. The Dialogue could help bridge that. Not by forcing alignment, but by making it easier to see where efforts overlap, where they conflict, and where there is room to collaborate. What would make it actually valuable is if people share what they are doing in practice, not just what they believe in. There is no shortage of principles. What is harder to find is how different countries or organizations are handling things like oversight, auditing, or harm response in real situations. That kind of exchange would make cooperation more concrete. It also has a role in who gets included. A lot of international cooperation spaces still centre the same actors. If this Dialogue creates more room for civil society, youth, and communities who are already dealing with the impacts of AI, it changes the conversation in a meaningful way. At the same time, I do not think full global alignment is realistic. But even some shared expectations around transparency or accountability would make cross-border coordination easier. Ultimately, its value depends on what happens after. If it leads to ongoing relationships, clearer coordination, and people leaving with a better sense of who is responsible for what, then it is doing something useful. If not, it risks becoming just another conversation that does not translate into action.

There's already a lot happening in this space. The issue is not a lack of initiatives, it's that they're scattered and not always connected. You've got things like the OECD AI Principles, UNESCO's Recommendation on the Ethics of AI, the Global Partnership on AI (GPAI), and more recently the UK-led AI Safety Summit process. On the technical and governance side, there's also work coming out of standards bodies and industry-led efforts. In Canada specifically, there's movement around AIDA and broader digital governance conversations. All of these are doing something useful, but they are not always speaking to each other. What the AI Dialogue can do is act as a connector across these spaces. Not by replacing them, but by helping map what already exists and where the gaps are. Right now, it's hard to see how these efforts fit together or where there is duplication. It could also create more space to translate between different groups. A lot of policy conversations stay high-level, while technical communities are working in more detail, and civil society is often reacting after the fact. Bringing those perspectives into the same conversation, in a more structured way, would make the work more grounded. Another piece is accessibility. Many of these initiatives are not easy to engage with unless you are already in those networks. The Dialogue could lower that barrier and make it easier for more actors, especially from underrepresented regions or sectors, to plug in. The added value, for me, is not new frameworks. It is coordination, clarity, and making sure the work that already exists actually connects and moves forward in a more coherent way.

Different stakeholders should not just be "included" in the Dialogue, they should have clear roles in shaping it. Governments can bring regulatory direction and commit to concrete next steps. Industry should be expected to share how systems are actually being built, deployed, and monitored, not just high-level commitments. Civil society and community organizations need space to bring lived experience and real-world impacts into the conversation, especially where harm is already happening. Researchers and technical experts can help translate what is feasible and where risks are emerging. Youth should not be treated as symbolic participants. They are already navigating these systems daily and should be engaged as contributors, not just consulted. In terms of format, smaller, focused discussions would work better than large, general sessions. The more open-ended the conversation, the easier it is to stay vague. The Dialogue should be structured around specific problems, like accountability mechanisms or harm response, with the expectation that each group produces something tangible. There should also be a mix of open and closed spaces. Open sessions allow for transparency and broader participation, while smaller working groups can support more honest conversations and actual problem-solving. Another important piece is continuity. This should not be a one-time event. There needs to be follow-up, whether through working groups, check-ins, or shared progress tracking, so that ideas do not just stay in the room. Finally, outcomes should be simple and usable. Short summaries, clear commitments, and accessible outputs matter more than long reports that few people will read. If the structure pushes people toward specificity and shared responsibility, the Dialogue will be much more useful.

A lot of the same voices still dominate AI governance, mostly governments, large tech companies, and a small group of well-resourced experts. What's missing are the people who are actually living with the impacts. Communities in the Global South are still underrepresented, especially in decision-making roles. They are often included late, or in consultation formats that don't carry real influence. The same goes for Indigenous communities, who bring very different perspectives on data, ownership, and governance that are not well reflected in current frameworks. Youth are also consistently underestimated. In my work, I see how deeply young people are already navigating AI-driven systems, from education to online spaces, but their input is often treated as symbolic rather than substantive. There's also a gap when it comes to people with disabilities and neurodivergent communities. AI systems can either improve access or create new barriers, but those perspectives are not consistently built into design or policy conversations. Frontline workers and community organizations are another group that gets overlooked. They are often the first to see how systems are actually working or failing in practice, but their insights rarely make it into global discussions. In terms of inclusion, it's not just about inviting people into the room. It's about shifting how participation works. That means resourcing participation, not expecting unpaid contributions, and creating formats where different kinds of knowledge are taken seriously, not just technical or policy expertise. It also means involving these groups earlier, when agendas are being set, not after decisions are already shaped. If inclusion is treated as a checkbox, nothing changes. If it is built into how decisions are made, it starts to shift who AI governance is actually for.

Most AI conversations feel static because they rely on panels and speeches. If the goal is real engagement, the format has to push people to actually work through problems together. One format that would help is scenario-based working sessions. Instead of abstract discussions, give mixed groups a concrete situation, like an AI harm case or a public sector deployment, and have them figure out what accountability or response should look like. It forces people to move from principles to decisions. Cross-sector breakouts would also be important. Not grouping people by role, but intentionally mixing government, industry, civil society, and technical folks in the same room. That is where you start to see where assumptions break down and where coordination is actually needed. Another useful approach is live policy or system "stress-testing." Take an existing framework or tool and walk through how it holds up under pressure. Where does it fail? Who is responsible? What is missing? That kind of exercise makes gaps very visible. There should also be space for community-led sessions, where affected groups set the agenda and others respond. Not just sharing experiences, but shaping what gets discussed. Short decision-focused sprints could help too. Give groups a limited amount of time to produce something concrete, like a draft accountability pathway or minimum standard. It keeps things from drifting into general conversation. And importantly, there should be structured follow-through. Even something simple like public trackers or commitments tied to these sessions would make the engagement feel less performative. If the format makes it hard to stay vague, you get better outcomes.

A few approaches stand out to me, mostly because they try to move beyond principles into something enforceable or usable. The EU AI Act is one example. It is not perfect, but the risk-based approach gives a clearer structure for how different types of AI should be governed, especially around high-risk systems. What's useful is that it starts to tie obligations to actual use cases, not just broad ideas about "responsible AI." Australia's eSafety Commissioner model is another one I keep coming back to. It creates a clear pathway for harm reporting and response, especially around online abuse. With the rise of AI-generated content, that kind of mechanism feels increasingly relevant. It shows what it looks like when accountability is tied to a specific body with real authority. On the technical side, algorithmic auditing and impact assessments are gaining traction. When done properly, they force organizations to examine how systems might create harm before and after deployment. The challenge is consistency and enforcement, but the approach itself is solid. I also think there is value in community-informed design and governance, even though it is less formalized. In my own work, I have seen how bringing in lived experience earlier changes how systems are shaped. It is slower, but it leads to better outcomes because it surfaces issues that would otherwise be missed. Finally, there are emerging multi-stakeholder initiatives that try to bring policy, technical, and community perspectives together. They are still evolving, but they point toward a more collaborative model of governance. None of these are complete solutions on their own. But together, they show a shift toward more practical, enforceable, and context-aware approaches to governing AI.