Independent Researcher in AI Governance and Ethics

A successful first Global Dialogue on AI Governance should produce more than a broad exchange of views. It should generate concrete institutional outcomes. First, it should establish a shared baseline for global AI governance in high-impact contexts: meaningful human control, human rights protection, transparency of authority, accountability, contestability, and independent oversight. Second, it should identify a practical roadmap for interoperability across governance approaches. Different legal and political systems will not converge on a single model, but they should be able to align around minimum safeguards, common terminology, and mechanisms for cross-border cooperation. Third, success would require explicit recognition that high-impact AI systems need post-deployment governance, not only pre-deployment assessment. Monitoring, incident reporting, auditability, and redress mechanisms should be treated as core governance functions. Fourth, the process should ensure that participation leads to traceable follow-through. Stakeholders should be able to see how their input informs priorities, structures, and future action. Finally, the Dialogue should send a clear signal that AI governance is not only about innovation management. It is also about preserving human agency, preventing harmful concentration of power, and ensuring that AI remains subordinate to human dignity, democratic accountability, and the public interest.

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

I selected these four priorities because, taken together, they form the core of a credible and enforceable global AI governance architecture. Safe, secure and trustworthy AI is essential because high-impact systems must not create unacceptable risks for individuals, institutions, or societies. Protection and promotion of human rights is fundamental because AI governance loses legitimacy if it does not safeguard dignity, liberty, equality, due process, and non-discrimination. Transparency, accountability, and human oversight are necessary to ensure that AI systems remain understandable, contestable, and governable in practice. Human responsibility cannot remain merely symbolic where AI shapes consequential decisions. Interoperability of governance approaches is equally important because AI systems, supply chains, and impacts are cross-border. A fragmented governance environment will weaken enforcement, create legal uncertainty, and reduce the ability of states and institutions to cooperate effectively. These priorities are mutually reinforcing. Rights without oversight are difficult to protect. Oversight without interoperability remains fragmented. Interoperability without trust and accountability risks becoming procedural rather than substantive. Together, these areas provide the strongest foundation for human-centered, globally usable, and institutionally credible AI governance.

Yes. One major cross-cutting issue is the gap between governance on paper and governability in practice. Many frameworks describe principles or compliance expectations, but less attention is given to whether AI systems remain auditable, contestable, and controllable once deployed in real operational environments. A second emerging issue is post-deployment drift and institutional dependency. Risks do not end at model release or market entry. They often emerge through changing workflows, automation bias, degraded human judgment, unequal impacts, and growing dependence on systems that institutions may no longer be able to question or replace easily. A third issue is concentration of compute, infrastructure, and model power. Governance discussions often focus on models and applications, but not enough on the structural concentration that can distort competition, weaken public oversight, and increase dependency of smaller states and lower-capacity institutions. A fourth issue is the need for stronger anti-weaponization and anti-domination safeguards. AI governance should address not only safety failures, but also coercive surveillance, automated repression, destabilizing military use, and the erosion of civic freedom. Finally, capacity-building should be treated as a governance issue, not a secondary development issue. Many jurisdictions will struggle not because they reject safeguards, but because they lack the technical, legal, and administrative capacity to implement them.

In Eastern Europe and comparable lower-capacity jurisdictions, the main AI governance challenge is not the absence of principles, but the gap between formal commitment and operational capacity. Important advances are visible: growing awareness of AI risks, stronger alignment with international discussions, and increasing recognition that human rights, accountability, and oversight must be part of AI adoption from the start. This creates an opportunity to build governance frameworks before high-impact deployment becomes deeply entrenched. However, several gaps remain significant. First, institutional capacity is uneven. Many public bodies and sectoral regulators still lack the technical expertise, staffing, and procedural tools needed to evaluate, audit, or challenge AI systems in practice. Second, legal and governance fragmentation creates uncertainty, especially where national frameworks, sectoral rules, procurement practices, and cross-border digital dependencies do not yet align. Third, human oversight often remains more declarative than operational: institutions may formally require accountability without ensuring that affected persons and decision-makers can actually understand, contest, or override AI-supported outcomes. For the broader sector, this creates a dual effect. The opportunity is that emerging regions can still design more human-centered and interoperable governance pathways without first inheriting large-scale institutional dependency. The risk is that lower-capacity jurisdictions may become rule-takers, technology-dependent, or vulnerable to opaque deployments shaped elsewhere. The most important priority is therefore to connect governance principles to real implementation capacity: training, auditability, documentation, procurement safeguards, redress mechanisms, and clear lines of human responsibility. Without that, advances in AI governance will remain formal progress, but not yet effective governability.

The AI Dialogue can play a unique role by serving as a bridge between high-level principles and practical international coordination. First, it can help establish a globally legible baseline for AI governance: meaningful human control, protection of human rights, transparency of authority, accountability, contestability, and independent oversight for high-impact systems. Second, it can reduce fragmentation by identifying areas where interoperability is possible across different legal and political systems. Full uniformity is unlikely, but states can still align around minimum safeguards, shared terminology, and mechanisms for cross-border cooperation. Third, the Dialogue can elevate issues that require collective action beyond national regulation alone, including cross-border deployment, concentration of technological power, compute and infrastructure dependency, and the risk of harmful or destabilizing uses of AI. Fourth, it can create a more balanced global process by giving structured voice to smaller states, lower-capacity jurisdictions, civil society, and public-interest experts, not only major powers and large technology firms. Finally, the Dialogue can strengthen legitimacy if it ensures that consultation leads to visible institutional outcomes. International cooperation becomes credible when stakeholder input is translated into priorities, coordination mechanisms, and follow-through that states and institutions can actually use. In this sense, the AI Dialogue should not function only as a forum for discussion. It should function as a mechanism for convergence: helping transform distributed concerns and diverse experiences into more coherent, rights-respecting, and operationally usable approaches to global AI governance.

The AI Dialogue should build upon and connect with existing international work rather than duplicate it. Important foundations already exist, including the UNESCO Recommendation on the Ethics of Artificial Intelligence, the OECD AI Principles, the Council of Europe Framework Convention on Artificial Intelligence, and international standards work through ISO/IEC JTC 1/SC 42. It should also connect with partnerships and expert networks that support policy coordination, technical exchange, and implementation across jurisdictions. The added value of the AI Dialogue should be threefold. First, it can serve as a coordination layer across initiatives that currently operate in parallel. Many existing frameworks are valuable, but policymakers and institutions still face fragmentation across legal, ethical, technical, and sector-specific workstreams. Second, it can help translate principles into governable practice by highlighting where convergence is most urgently needed: human oversight, auditability, redress, interoperability, post-deployment monitoring, and safeguards for high-impact public-interest uses. Third, it can provide broader legitimacy by ensuring more inclusive participation from smaller states, lower-capacity jurisdictions, civil society, and public-interest communities whose perspectives are often underrepresented in global standard-setting. The Dialogue would be most useful not as another standalone framework, but as a connective mechanism: aligning existing instruments, identifying unresolved governance gaps, and turning international consultation into more coherent and institutionally actionable cooperation.

Different stakeholders should contribute in ways that reflect their distinct institutional roles and practical knowledge. Governments can identify regulatory needs and cooperation gaps. Civil society can highlight rights impacts, exclusion risks, and accountability concerns. Academia can contribute independent analysis and comparative evidence. Technical communities can clarify feasibility, limitations, and implementation tradeoffs. The private sector can provide insight into deployment realities, supply chains, and compliance constraints. Frontline professional communities can show how AI actually changes decisions in practice. For the AI Dialogue to benefit from this diversity, its format should be structured rather than purely conversational. First, it should combine open consultation with thematic working tracks that produce focused outputs. Second, it should distinguish clearly between high-level principles, implementation challenges, and unresolved governance gaps. Third, contributions should be synthesized transparently, with public summaries showing which concerns were raised, how they were grouped, and how they informed priorities. The Dialogue should also include mechanisms that support meaningful participation across regions and capacities: multilingual access, hybrid participation, advance publication of questions, concise background papers, and formats that do not privilege only well-resourced institutions. Finally, the structure should not end with consultation. It should include traceable follow-through: published outcome documents, identified areas for convergence, and clear links between stakeholder input and subsequent institutional action. A successful Dialogue will depend not only on who participates, but on whether participation is translated into legible priorities, durable mechanisms, and accountable follow-up.

Several perspectives remain underrepresented in global AI governance discussions. First, smaller states and lower-capacity jurisdictions are often present formally but underweighted in agenda-setting. Yet they face distinct challenges: administrative constraints, technology dependence, procurement vulnerability, and limited regulatory capacity. Second, frontline professional communities are often overlooked. Healthcare workers, teachers, judges, public administrators, social workers, and labor representatives can identify how AI changes real decisions, workflows, and accountability structures in practice. Third, affected communities are still too rarely included in a structured way. People exposed to automated decision-making, surveillance, exclusion, discrimination, or unequal digital access should not appear only as abstract beneficiaries of governance, but as contributors to it. Fourth, public-interest researchers, independent experts, and civil society actors from the Global South remain underrepresented relative to large states, major firms, and highly resourced institutions. Inclusion should therefore be designed intentionally. This means reserved speaking opportunities, regional balance in agenda-setting, multilingual participation, travel and access support where needed, targeted outreach to underrepresented sectors, and consultation formats that allow written and asynchronous input. Most importantly, inclusion should not be symbolic. These voices should be visible not only in participation lists, but in synthesis documents, priority-setting, and the design of follow-up mechanisms. Global legitimacy depends not only on openness, but on whether diverse experience is allowed to shape the architecture of governance itself.

Meaningful engagement will require formats that move beyond one-way statements and allow structured interaction across different stakeholder groups. First, the Dialogue should combine plenary sessions with small, moderated working groups organized around concrete governance problems rather than only broad themes. This makes it easier to compare practical experiences across regions and sectors. Second, scenario-based discussions would be highly valuable. Participants could respond to shared cases involving high-impact AI in areas such as health, public administration, education, labor, or critical infrastructure. This would ground abstract principles in operational realities. Third, the process should include written asynchronous input before and after live sessions. This would improve participation from different time zones, smaller institutions, and stakeholders who may not be best represented in fast-moving verbal exchanges. Fourth, synthesis workshops should be used to identify where there is convergence, where disagreement remains, and what follow-up mechanisms are needed. These workshops should produce short public summaries so participants can see how discussion is being translated into governance priorities. Fifth, the Dialogue should include cross-stakeholder roundtables that intentionally mix governments, civil society, academia, technical experts, frontline professionals, and affected communities, rather than separating them into isolated tracks. Finally, dynamic engagement depends on visible follow-through. Participants should not only be invited to speak; they should be able to trace how their contributions shape agendas, outputs, and future institutional work. The most effective format is therefore hybrid, multilingual, case-based, and synthesis-oriented: open enough for inclusion, but structured enough to produce actionable outcomes.

Useful examples already exist. The UNESCO Recommendation on the Ethics of Artificial Intelligence provides a rights-based normative baseline for governance. The OECD AI Principles offer a widely recognized policy foundation, while the UK's Algorithmic Transparency Recording Standard shows how public-sector documentation and accountability can be operationalized in practice. The Council of Europe Framework Convention on Artificial Intelligence is a concrete legal model linking AI governance to human rights, democracy, and the rule of law. Technical interoperability can be further strengthened through international standards work, including ISO/IEC JTC 1/SC 42. In practice, effective governance is also supported by measures such as impact assessments, procurement safeguards, documentation requirements, audit trails, post-deployment monitoring, incident reporting, and redress mechanisms for affected persons. The broader lesson across these examples is that governance works best when principles are connected to documentation, oversight, reviewability, and institutional capacity. The AI Dialogue can add value by identifying such practices, comparing them across jurisdictions, and promoting more interoperable, human-centered approaches that remain workable for both high-capacity and lower-capacity states.