RBX Labs

A successful first Global Dialogue would move beyond broad principles and produce a practical, shared agenda for implementation. From RBX Labs' perspective, success would mean three things: First, clearer common language for what safe, trustworthy, and accountable AI looks like in real systems. Second, stronger channels for exchanging implementation practices across countries and sectors and Third, concrete priorities for capacity-building so that smaller teams, public institutions, and developing markets are not left behind as governance expectations rise. The Dialogue was created to support international cooperation, share best practices and lessons learned, and enable open, transparent, and inclusive discussion on AI governance. A successful first session should show how that mandate translates into action. For RBX Labs, this is not abstract. In work such as Network Guardian, a pre-connection trust layer for public Wi-Fi, and in AI trust, ranking, and moderation systems built at Kampd, the governance questions show up at the workflow level: what signals are used, where decisions are made, how confidence is communicated, and when humans stay in the loop. A strong outcome from the Dialogue would be recognition that governance must address not only frontier models, but also the smaller, embedded systems that shape everyday decisions in infrastructure, communities, and public-facing products. That would make the process more useful for practitioners and more relevant globally.

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

These four areas best reflect where RBX Labs can contribute practical experience. Safe, secure and trustworthy AI is central because many harms emerge not from abstract model behavior alone, but from how AI systems operate inside real workflows. In Network Guardian, for example, the key question is how to provide trustworthy signals before a user or application commits to a risky network. In systems work such as Kampd, trust also depends on how ranking, moderation, and policy decisions are structured and monitored in production. Transparency, accountability, and human oversight matter because real systems need visible decision points, understandable signals, and clear ownership. In practice, this means making confidence, constraints, and escalation paths explicit rather than assuming good outcomes from model performance alone. Interoperability of governance approaches is also important because fragmented rules make it harder for teams building across regions and product surfaces to implement governance consistently; several recent analyses point to interoperability as a core challenge for effective AI governance. Finally, AI capacity-building is essential because governance only works if organizations, especially smaller teams and less-resourced ecosystems, can actually understand and operationalize it. RBX Labs is particularly interested in helping bridge the gap between governance principles and deployable practice.

Yes. One important cross-cutting issue is workflow-level governance: how AI behaves inside real operational systems once models are connected to data sources, interfaces, ranking logic, rules, and human decision-makers. Many governance discussions remain focused on models, providers, or high-level principles, but the real-world impact of AI is often determined by how these systems are assembled and used in context. Research on global AI governance has also highlighted the disconnect between high-level governance design and grassroots or implementation realities. A second emerging issue is pre-action trust. In many real environments, the most important governance question is not what happens after a system fails, but what information and safeguards exist before a user, team, or agent acts. That is especially relevant in public infrastructure, safety-sensitive workflows, and systems that rely on uncertain external environments. A third issue is governance for embedded and smaller-scale AI systems. A large share of real-world AI impact now comes from workflow tools, copilots, ranking systems, trust layers, and operational automation, not only from frontier general-purpose models. Governance frameworks will be stronger if they account for these systems explicitly, because that is where adoption, accountability, and public trust are often won or lost.

In US & Canada and in the private-sector environments we work with, the biggest governance gap is no longer awareness of AI, but the ability to operationalize it responsibly. Organizations are moving ahead with adoption, but they are doing so amid unclear or evolving legal expectations, fragmented governance approaches, and uneven internal capacity. In Canada, businesses themselves report uncertainty around legal and regulatory requirements as a major barrier to implementation, while recent federal and provincial developments continue to signal rising expectations around risk management, transparency, and responsible deployment. A second challenge is that many governance discussions still operate at the level of models or policy statements, while the real risks and opportunities often emerge at the workflow level. In sectors like digital infrastructure, community platforms, and SME software, the important questions are practical: what data is used, where decisions are made, how confidence is communicated, and when humans remain in the loop. We see this directly in work such as Network Guardian, where trust must be established before users or systems connect to public networks, and in Kampd, where ranking, moderation, and trust signals shape everyday user experience. A major opportunity is that organizations that build governance into actual system behavior, not just high-level policy, can move faster with more trust and clearer accountability. This is increasingly recognized as a competitive advantage in responsible AI adoption. For US & Canada specifically, another important tension is between strong AI ambition and uneven adoption capacity. US & Canada have made SME AI adoption a policy priority, but OECD work notes that SME adoption still lags larger firms, even as use of off-the-shelf AI tools rises. That creates both a challenge and an opportunity: smaller organizations need clearer workflows, better integration guidance, and practical AI literacy, not just access to tools. The biggest opportunity is to turn governance into an enabler of adoption by making it concrete, interoperable, and usable for teams that are actually deploying systems in practice.

The AI Dialogue can play a critical role in bridging the gap between high-level governance principles and real-world implementation. Today, many international efforts define what responsible AI should look like, but there is less alignment on how those principles translate into systems that operate across different countries, infrastructures, and use cases. A key role for the Dialogue is to create a shared space where governments, practitioners, and industry can exchange implementation-level practices. This includes how trust is established in real systems, how decisions are made and communicated, and how accountability is maintained when AI is embedded in workflows. From RBX Labs' perspective, this is where governance becomes actionable. The Dialogue can also support interoperability by identifying common building blocks across jurisdictions. Rather than forcing uniform regulation, it can help align on core mechanisms such as trust signaling, human oversight, and evaluation standards, allowing systems to operate consistently across borders. Finally, the Dialogue can act as a bridge between advanced and emerging ecosystems by ensuring that governance approaches are not only robust but also usable. This includes sharing practical models, templates, and examples that smaller organizations and public agencies can adopt without significant overhead. By focusing on implementation, interoperability, and accessibility, the AI Dialogue can move international cooperation from alignment in principle to alignment in practice.

The AI Dialogue should build upon existing global frameworks and initiatives that have already established strong foundations in principles, standards, and capacity-building. This includes efforts such as the OECD AI Principles, UNESCO's Recommendation on the Ethics of AI, and regional regulatory developments that are shaping expectations around safety, transparency, and accountability. It should also connect with implementation-focused initiatives such as ITU AI for Good programs, sandbox environments, and other multi-stakeholder platforms that are testing real-world applications. While these efforts have advanced a shared understanding of responsible AI, they often operate in parallel. The added value of the AI Dialogue lies in its ability to connect these initiatives by surfacing practical implementation lessons and enabling greater coherence, interoperability, and complementarity across frameworks, rather than creating additional layers of governance. From a practitioner perspective, including work at RBX Labs, a key gap today is the limited sharing of system-level insights. For example, how trust signals are designed, how human oversight is embedded into workflows, or how systems are evaluated before deployment. These are critical to real-world governance but are not consistently captured across initiatives. The AI Dialogue can create a structured feedback loop between policy, standards, and implementation by bringing these insights into a shared global forum. By identifying common patterns that work across contexts and feeding them back into existing frameworks, it can help move governance from alignment in principle to alignment in practice.

Different stakeholders should contribute in ways that reflect their actual strengths. Governments can articulate policy priorities and areas where international cooperation is most needed. Private sector participants can share implementation lessons from deploying AI systems in real products and workflows. Academia and technical communities can contribute evidence, evaluation methods, and emerging research. Civil society can help ensure the Dialogue remains grounded in rights, inclusion, and public accountability. International organizations can connect existing initiatives and reduce fragmentation. To make this effective, the Dialogue should combine high-level plenary sessions with smaller, structured thematic discussions. The plenaries are useful for setting shared priorities, but the most valuable exchanges are likely to happen in breakout formats where participants can discuss concrete use cases, governance gaps, and implementation lessons in more detail. Each thematic track should include a mix of governments, technical practitioners, civil society, and researchers rather than grouping similar actors together. A useful structure would include short scene-setting interventions, followed by moderated discussion, and then a practical synthesis of takeaways and possible cooperation actions. The Dialogue should also create mechanisms for written inputs, case-based submissions, and follow-up between annual sessions, so that it becomes a continuing process rather than a one-off event. From the perspective of RBX Labs, the most important design principle is that the Dialogue should make room for system-level lessons from deployment, not only high-level principles. That is where governance becomes usable in practice.

Several voices remain underrepresented in global AI governance discussions. One is smaller implementation-focused teams: startups, SMEs, civic technologists, and applied AI builders who are not shaping foundation models but are responsible for how AI is actually embedded into workflows, products, and public-facing systems. These actors often face the real operational tradeoffs of governance, but their lessons are not consistently reflected in global conversations. A second underrepresented group is practitioners from developing and lower-capacity ecosystems who work with constrained infrastructure, smaller datasets, and different deployment realities. Their experience is essential for ensuring governance frameworks are globally relevant rather than optimized only for large, well-resourced institutions. A third underrepresented perspective is frontline operators inside institutions and companies: product managers, operations leads, trust and safety teams, and public service implementers. These are the people who often decide how AI systems are configured, monitored, and escalated in practice, yet their role is rarely foregrounded in governance forums. These groups can be included through targeted participation pathways rather than open invitations alone. That means reserved speaking opportunities in thematic sessions, practical case submission tracks, regional consultations, and support for remote participation. The Dialogue should also value implementation evidence alongside formal policy contributions. At RBX Labs, one recurring lesson is that governance often succeeds or fails at the workflow level. Including more voices from real deployment contexts would make the Dialogue more grounded, more inclusive, and more useful.

The most effective engagement formats will be those that move beyond statements and toward shared examination of real governance problems. One strong format would be case-based roundtables, where participants analyze a concrete AI deployment scenario and discuss how governance issues such as safety, oversight, accountability, and interoperability arise in practice. This would make discussions more grounded and comparable across sectors. Another useful format would be "implementation clinics" or practitioner labs, where governments, technical experts, companies, and civil society can examine how a principle such as transparency or human oversight is translated into an actual system. These sessions would be especially valuable for surfacing practical lessons that do not usually emerge in plenary dialogue. A third option is a "dialogue of dialogues" format that brings together representatives from existing initiatives, standards bodies, regional frameworks, and applied projects to compare where they converge, where they differ, and what practical lessons can be shared. This would help reduce duplication and strengthen complementarity. Interactive breakout sessions should also be designed to mix stakeholders intentionally, rather than separating them by category. Short written prompts, moderated problem-framing, and clear synthesis at the end of each session would help keep exchanges substantive. From an RBX Labs perspective, the most valuable formats are those that let participants test governance ideas against real systems, workflows, and constraints. That is where dynamic engagement becomes meaningful and where the Dialogue can produce lessons that are useful beyond the room.