Ostrich AI Solutions & Integrated Systems Pvt. Ltd.

A successful first Global Dialogue should move beyond broad principles and produce practical alignment on how AI governance can be made operational across jurisdictions. In particular, success would mean three things. First, the Dialogue should identify a shared baseline of governance outcomes that can be recognized across legal and regulatory systems, even where laws differ. These include safety, accountability, traceability, human oversight, and meaningful control over how data and models are used. Second, it should help translate governance from policy language into implementable technical and organizational controls. Many institutions, especially in regulated sectors, do not struggle with understanding that AI should be governed; they struggle with how to operationalize that governance in real deployment environments. The Dialogue would be valuable if it encourages frameworks that are verifiable, auditable, and enforceable in practice. Third, it should create an inclusive path for capacity-building so that governance is not limited to a small group of highly resourced actors. Smaller companies, public institutions, and emerging markets need practical guidance, common vocabularies, and interoperable approaches they can adopt without excessive complexity. A strong first outcome would therefore be a clear foundation for continued international cooperation: one that is principles-based, technically grounded, and realistic enough to support deployment in the real world.

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

These four areas are closely connected and, in my view, deserve urgent attention because they determine whether AI governance can work in practice rather than remain purely aspirational. Safe, secure and trustworthy AI is foundational, especially as AI systems are increasingly deployed in sensitive and regulated environments. Trust cannot depend only on stated intentions; it must be supported by technical, procedural, and institutional safeguards. Interoperability of governance approaches is equally important because organizations increasingly operate across borders. While legal systems differ, many governance objectives overlap. International dialogue should therefore focus on how to create compatible operational approaches rather than assume a single uniform model. Transparency, accountability, and human oversight are essential because they make responsibility visible and enforceable. Without auditability, explainable decision pathways, and clearly assigned accountability, governance becomes difficult to verify. AI capacity-building is critical to ensure that governance does not become concentrated among only the most advanced or well-resourced actors. Policymakers, enterprises, technical teams, and public institutions all need practical guidance and implementation support. Taken together, these priorities can help bridge the gap between high-level governance principles and real-world deployment.

Yes. One important cross-cutting issue is the need to distinguish between governance principles and governance enforceability. Many current discussions focus on what responsible AI should look like, but less attention is given to whether those requirements can actually be enforced through system design, operational controls, and audit mechanisms. A second issue is the governance of AI deployment infrastructure itself. Governance does not apply only to models and outputs; it also applies to data access, execution environments, logging, key management, jurisdictional controls, and the handling of third-party components. These infrastructure-layer questions become especially important in regulated sectors. A third emerging issue is purpose-bound use and context integrity. AI governance should pay more attention to whether data, models, and outputs are used only for authorized and disclosed purposes, particularly when systems are reused across multiple workflows or entities. Finally, international discussions should pay greater attention to practical adoption by smaller institutions and emerging markets. Governance frameworks that are too abstract, expensive, or difficult to implement may unintentionally widen the global governance gap.

In the enterprise and regulated-sector context, the main governance gap is not a lack of principles but a lack of practical, interoperable implementation. Many organizations in sectors such as financial services, healthcare, telecom, and the public sector are under pressure to adopt AI, yet they face uncertainty about how to do so in a way that is secure, auditable, and compliant across jurisdictions. One major challenge is fragmentation. Different legal and regulatory frameworks use different terminology, thresholds, and compliance expectations. Even where governance goals are similar, organizations struggle to translate them into a consistent operational model. This slows adoption, increases compliance cost, and creates hesitation among risk, legal, and technical teams. A second challenge is the gap between governance requirements and technical architecture. In practice, organizations need enforceable controls around data access, model usage, audit logging, human oversight, and jurisdiction-sensitive deployment. Where these controls are weak or unclear, trust in AI deployment remains limited. At the same time, this challenge creates a major opportunity. There is growing demand for governance approaches that are practical, verifiable, and capable of working across borders and sectors. In my region and sector, this opens space for international cooperation on common governance baselines, implementation frameworks, and capacity-building for both institutions and technical teams. If addressed well, AI governance can become an enabler of adoption rather than a brake on innovation. Better alignment can help organizations deploy AI more confidently, especially in environments where trust, accountability, and compliance are essential.

In my view, the AI Dialogue can play a useful role by helping close the gap between broad AI governance principles and the realities of implementation across jurisdictions. Many countries and sectors are working toward similar goals — safety, accountability, transparency, human oversight, and respect for rights — but they often approach them through different legal, regulatory, and institutional models. The Dialogue can help by identifying where meaningful convergence already exists and where greater interoperability is possible. That matters because organizations increasingly develop, deploy, and use AI across borders, while governance obligations remain fragmented. International cooperation becomes more effective when it focuses not only on shared values, but also on practical ways to make those values operational. A particularly valuable role for the Dialogue would be to encourage discussion around implementation: common vocabularies, practical governance baselines, and examples of enforceable technical and organizational controls. In many cases, the challenge is no longer recognizing that AI should be governed, but understanding how governance can be translated into auditable, accountable systems in practice. The Dialogue can also strengthen cooperation by widening participation. Emerging markets, smaller enterprises, and technical practitioners should not be limited to commenting on governance after it is framed elsewhere. Their implementation experience is essential if international AI governance is to be realistic, inclusive, and durable.

In my view, the AI Dialogue should build on existing international frameworks and multistakeholder efforts that have already helped establish common governance foundations, including work around AI ethics, trustworthy AI, standards, and responsible innovation. The core principles are not the main gap anymore. The bigger challenge is fragmentation across policy, technical, and implementation communities. That is where the AI Dialogue can add real value. It should not duplicate what already exists, but connect those efforts more effectively. In particular, it can help bridge high-level governance language with the realities of deployment, especially in cross-border and regulated environments where organizations need workable approaches rather than abstract alignment. The Dialogue can be especially useful if it focuses on where practical convergence is possible: shared governance outcomes, interoperable approaches, and implementation lessons that can travel across jurisdictions even when legal systems differ. It can also help surface the operational questions that often receive less attention, such as auditability, traceability, accountable human oversight, and the enforceability of governance controls within real systems and workflows. I also believe the Dialogue would benefit from stronger participation by actors dealing with live deployment constraints, including technical practitioners, enterprises, and institutions in emerging markets. That would make international cooperation more grounded, more inclusive, and more useful in practice.

Different stakeholders should contribute not only by presenting positions, but by bringing specific implementation experience, risks, and lessons from their contexts. Governments can contribute regulatory priorities and public-interest objectives. Technical experts can clarify what is feasible, auditable, and enforceable in practice. Civil society can surface rights-based and societal concerns. Private sector participants can provide operational insight from real deployment environments. Academia can help frame evidence, evaluation methods, and longer-term implications. To make this useful, the Dialogue should combine plenary discussions with smaller, structured working sessions organized around concrete governance questions rather than broad themes alone. For example, sessions could focus on interoperability, accountability in deployment, capacity-building, or governance in high-impact sectors. I would also recommend using a format that distinguishes between principle-level discussion and implementation-level discussion. Too often these are blended together, which leads to repetition and limited progress. A more effective structure would allow one track to focus on shared governance goals and another to focus on practical pathways, constraints, and examples. Finally, the Dialogue should allow written submissions, breakout discussions, and post-event synthesis that clearly captures areas of convergence, disagreement, and open questions. That would make participation more meaningful and ensure the Dialogue produces cumulative value rather than a one-time exchange of statements.

In my view, global AI governance discussions still underrepresent those dealing with implementation under real constraints. This includes technical practitioners responsible for deployment, compliance and risk teams inside operating institutions, smaller enterprises, public-sector implementers, and stakeholders from emerging markets who are adapting governance expectations with fewer resources and different institutional realities. There is also not enough representation from sectors where AI adoption is sensitive and heavily conditioned by trust, accountability, and regulatory obligations, such as financial services, healthcare, telecom, education, and public administration. These sectors often face practical governance problems earlier and more sharply than others. Another underrepresented perspective is that of organizations working at the infrastructure and workflow layer, where governance requirements must actually be translated into system behavior. Global discussions often focus heavily on model behavior and principles, but less on whether governance can be enforced operationally. These voices can be included by lowering the barriers to participation and moving beyond invitation-based visibility. The Dialogue should support written contributions, regional consultations, implementation-focused roundtables, and structured inclusion of smaller actors, not only major governments and large technology firms. It should also actively include participants from the Global South, multilingual communities, and institutions outside the usual governance circles. A governance conversation is incomplete if it is dominated by those with the most visibility rather than those dealing most directly with implementation.

The most effective formats would be those that move the discussion from prepared statements to structured problem-solving. One useful approach would be scenario-based sessions where diverse stakeholders respond to a shared governance challenge, such as cross-border AI deployment, high-impact decision-making, incident response, or governance of third-party models and data flows. This would reveal where perspectives genuinely converge or differ in practice. Another strong format would be moderated implementation labs or policy-to-practice workshops. These could bring together policymakers, technical experts, civil society, and deployers to examine how broad governance goals translate into actual controls, processes, and accountability structures. That would make the Dialogue more concrete and less rhetorical. Short multi-stakeholder response rounds could also work well, where each participant addresses the same narrowly framed question with strict time limits. This creates sharper comparison than long general remarks and helps surface practical insights quickly. In addition, regional and sector-focused breakout sessions would be valuable, especially for capturing differences in institutional capacity, legal context, language needs, and deployment realities. A final synthesis session should then identify cross-cutting lessons, tensions, and implementation priorities. In my view, the best engagement formats are those that make participants do more than speak. They should help stakeholders compare assumptions, test governance ideas against reality, and leave with clearer shared understanding.

In my view, the most effective approaches to AI governance are those that combine policy intent with operational enforceability. At the policy level, risk-based governance approaches are useful because they help distinguish between low-risk and high-impact uses rather than treating all AI systems the same. This makes governance more practical and proportionate. Similarly, governance frameworks that emphasize accountability, transparency, human oversight, safety, and redress provide a strong baseline when they are translated into real processes. At the practice level, some of the most useful measures are impact assessments, model and dataset documentation, audit logging, access controls, role-based approvals, and clear escalation paths for incidents or high-risk decisions. Human oversight is most effective when it is designed into workflows rather than added as a formal requirement on paper. A strong governance approach should also address the infrastructure layer. In practice, effective governance often depends on whether institutions can enforce controls over data access, model use, execution environments, traceability, and jurisdiction-sensitive deployment. This is especially important in regulated or cross-border settings. Another good practice is to build interoperability into governance from the start. Many organizations now operate across multiple legal and regulatory environments, so approaches that map shared governance outcomes across jurisdictions are often more practical than narrowly siloed compliance models. Overall, the most promising examples are not only principle-based, but verifiable, auditable, and usable in real deployment conditions.