THS Soluções Institucionais Integradas

A successful first Global Dialogue on AI Governance would move beyond high-level principles toward actionable institutional guidance for governments, particularly at the national and local levels. While global frameworks are essential, public administrations urgently need practical roadmaps to translate governance principles into operational decision processes. Three outcomes would be especially valuable. First, the establishment of shared baseline standards for institutional readiness — including risk assessment protocols, accountability mechanisms, and human oversight structures — before large-scale AI deployment in public services. Second, mechanisms for international cooperation that support capacity-building in developing countries and subnational governments, where implementation gaps are most pronounced. Third, the creation of a continuous, multi-stakeholder platform that connects policymakers, technical experts, civil society, and public managers to exchange applied experiences and lessons learned. Equally important is recognizing that AI governance is not solely a technological issue but an institutional design challenge. Governments require structured decision architectures to ensure that AI systems are deployed coherently, transparently, and responsibly. If the Dialogue can produce concrete guidance, foster trust across stakeholders, and accelerate institutional preparedness rather than only normative discussions, it will represent a meaningful step toward safe and equitable AI adoption worldwide.

• AI capacity-building
• Safe, secure and trustworthy AI
• Interoperability of governance approaches

These priorities reflect the foundational conditions required for responsible AI adoption in the public sector. Transparency, accountability, and human oversight are essential to maintain democratic legitimacy and public trust, especially when automated systems influence rights, services, or resource allocation. Without clear lines of responsibility and auditable decision processes, governments risk amplifying opacity rather than improving governance. AI capacity-building is equally urgent. Many governments, particularly at subnational levels, lack the technical, institutional, and organizational capabilities needed to evaluate, procure, and manage AI systems effectively. Strengthening these capacities is a prerequisite for equitable global adoption. Safe, secure, and trustworthy AI depends not only on technical robustness but also on governance frameworks that systematically assess risks, monitor impacts, and enable corrective action. This requires structured institutional processes rather than ad hoc implementation. Finally, interoperability of governance approaches is crucial in an interconnected world. Divergent regulatory models can create fragmentation, barriers to cooperation, and uneven protection standards. Aligning principles, practices, and technical norms across jurisdictions will facilitate international collaboration while respecting national contexts. Together, these areas address both the technical and institutional dimensions of AI governance, emphasizing preparedness, accountability, and coordination.

One critical cross-cutting issue insufficiently captured by the listed themes is institutional decision readiness. Many governments are accelerating AI adoption while their underlying decision-making processes remain fragmented, opaque, or weakly structured. Without coherent internal governance, even well-designed AI systems can produce inconsistent, unaccountable, or high-risk outcomes. Public administrations need frameworks that define how decisions are prepared, justified, documented, and reviewed before automation occurs. This includes systematic comparison of alternatives, explicit criteria for action, structured risk mapping, and clear accountability chains. In the absence of such decision architectures, AI may amplify existing inefficiencies or biases rather than improve performance. Another emerging issue is the operational gap between global principles and local implementation. Cities and regional governments often face the most immediate pressures to deploy AI but have the least institutional capacity to do so safely. Mechanisms tailored to subnational contexts are therefore essential. Additionally, long-term institutional resilience should be considered. AI systems will evolve rapidly, and governance mechanisms must be adaptable, auditable, and capable of continuous learning. Addressing decision readiness alongside technical governance would help ensure that AI strengthens public institutions rather than exposing them to new systemic risks.

Governance gaps in AI are particularly acute in middle-income countries and decentralized systems such as Brazil, where national strategies coexist with highly autonomous subnational governments that face immediate pressure to modernize public services. The result is an uneven landscape: advanced digital initiatives in some jurisdictions alongside limited institutional capacity in others. The most significant challenge is not the absence of regulation, but the mismatch between technological adoption and institutional preparedness. Many public organizations lack structured processes for risk assessment, procurement evaluation, accountability allocation, and long-term oversight of AI systems. This creates exposure to operational failures, legal uncertainty, and erosion of public trust, especially in sectors such as social protection, justice, public procurement, and urban services. Capacity constraints are particularly severe at the municipal level, where financial resources, technical expertise, and governance tools are often limited. Without coordinated support, AI deployment risks reinforcing regional inequalities rather than reducing them. Fragmented governance approaches also hinder interoperability, data sharing, and collaborative innovation across jurisdictions. At the same time, the opportunities are substantial. Emerging economies can leapfrog legacy systems by adopting AI within modern digital infrastructures, provided that robust governance frameworks are in place. Regional cooperation and knowledge exchange can accelerate this process, enabling scalable solutions adapted to local contexts. Ultimately, the challenge is to ensure that AI strengthens state capability rather than outpacing it. Closing governance gaps requires not only technical safeguards but also institutional reforms that enhance transparency, accountability, and decision quality across all levels of government.

The AI Dialogue can play a pivotal role as a neutral convening platform capable of bridging fragmented global efforts on AI governance. In a context where regulatory approaches, technical standards, and policy priorities vary widely across regions, the Dialogue can foster convergence without imposing uniformity, enabling cooperation while respecting national sovereignty and developmental differences. One critical contribution would be facilitating structured exchange between countries at different stages of digital maturity. Many developing and middle-income countries face similar implementation challenges but lack channels to systematically share practical experiences, policy tools, and lessons learned. The Dialogue could support peer learning networks, technical cooperation, and capacity-building initiatives tailored to diverse institutional contexts. Additionally, the platform can help align public sector priorities with technical communities, private sector actors, and civil society, reducing the disconnect that often exists between normative frameworks and operational realities. By encouraging transparency, trust-building, and inclusive participation, the Dialogue can mitigate risks of technological fragmentation and governance asymmetries. Another important role is to anticipate emerging risks and coordinate collective responses, particularly in areas where unilateral action would be insufficient, such as cross-border data flows, AI safety standards, and systemic impacts on labor markets and democratic institutions. Ultimately, the Dialogue's value lies in its ability to transform parallel national efforts into a coherent global conversation, accelerating responsible AI adoption while strengthening international trust and cooperation.

The AI Dialogue should build upon existing multilateral, regional, and technical initiatives that have already advanced principles, standards, and policy frameworks for AI governance. Notable examples include the OECD AI Principles, UNESCO's Recommendation on the Ethics of Artificial Intelligence, regional regulatory efforts such as the European Union's AI Act, and ongoing work within organizations such as the Council of Europe, the World Bank, and the International Telecommunication Union. Numerous national strategies and public sector innovation programs also provide valuable implementation experience. However, these initiatives often operate in parallel, with limited coordination across institutional domains and geographic regions. The added value of the AI Dialogue would be to create a high-level platform for synthesis, interoperability, and political alignment among these efforts, reducing duplication and fragmentation. In particular, the Dialogue could facilitate connections between normative frameworks and operational practice, ensuring that principles translate into actionable guidance for governments. This includes sharing methodologies for risk assessment, procurement, oversight, and evaluation of AI systems in public administration. Another key contribution would be amplifying perspectives from the Global South and subnational governments, which are frequently underrepresented in global standard-setting processes despite facing urgent implementation pressures. By integrating these voices, the Dialogue can promote more equitable and context-sensitive governance models. Ultimately, the Dialogue can serve as a bridge between existing initiatives, transforming a landscape of dispersed efforts into a coordinated ecosystem that supports safe, inclusive, and sustainable AI development worldwide.

Different stakeholders can contribute most effectively if the AI Dialogue is structured as a multi-layered process that combines high-level policy discussion with practical, implementation-oriented exchanges. Governments, international organizations, private sector actors, academia, and civil society each possess distinct expertise that should be mobilized in complementary ways. Governments can share regulatory approaches, national strategies, and lessons learned from public sector deployment. The private sector can contribute technical knowledge, innovation trends, and insights into operational constraints. Academic institutions and research organizations can provide independent evidence, foresight analysis, and methodological rigor. Civil society can highlight societal impacts, ethical considerations, and risks to vulnerable populations. To harness these contributions, the Dialogue should include a mix of plenary sessions for strategic alignment, thematic working groups focused on concrete issues, and technical roundtables that produce actionable outputs. Structured mechanisms for written submissions, case studies, and peer learning exchanges would allow participation beyond those physically present. Importantly, the process should not be a one-off event but an ongoing platform with follow-up mechanisms, knowledge repositories, and opportunities for continuous engagement. Digital participation tools can expand accessibility and ensure representation from countries with limited travel capacity. By combining inclusiveness with structured outputs, the AI Dialogue can move from general debate toward practical cooperation, enabling stakeholders to co-create solutions and support responsible AI governance across diverse institutional contexts.

Global discussions on AI governance often underrepresent actors operating at the front lines of implementation. Subnational governments — including cities, regions, and local administrations — face immediate pressures to deploy AI in public services but rarely participate meaningfully in global standard-setting processes. Their practical experience with procurement, service delivery, and citizen interaction is essential for translating principles into workable governance models. Small and medium-sized enterprises, public sector innovators, and organizations from the Global South are also frequently underrepresented. These actors often lack the resources, networks, or institutional channels to engage in international forums, despite being key drivers of context-sensitive innovation. Similarly, marginalized communities and populations most affected by automated decision-making may have limited voice in shaping governance frameworks. To address these gaps, the Dialogue should adopt inclusive participation mechanisms such as targeted outreach, funding support for participation, regional consultations, and digital engagement platforms that reduce barriers to entry. Partnerships with local government associations, development agencies, and regional organizations can help identify and mobilize relevant stakeholders. Additionally, incorporating structured case studies and field-based evidence would allow practitioners to contribute without requiring formal diplomatic representation. Ensuring linguistic diversity and accessible formats is also critical to broaden participation. By actively integrating these perspectives, the AI Dialogue can produce governance approaches that are not only normatively robust but also operationally feasible across diverse socio-economic contexts

Innovative engagement formats should prioritize interactive, problem-solving approaches rather than purely declaratory discussions. Traditional conference models often produce high-level statements but limited operational insight. To foster meaningful engagement, the AI Dialogue could incorporate formats that simulate real governance challenges and encourage collaborative solutions. One promising approach is policy labs or scenario-based exercises, where diverse stakeholders work together on concrete case studies, such as AI deployment in public services, cross-border data governance, or crisis response. These structured simulations can reveal trade-offs, institutional constraints, and practical implementation barriers that may not surface in formal speeches. Another effective format would be multi-stakeholder design sprints focused on producing tangible outputs, such as draft guidelines, risk assessment frameworks, or implementation roadmaps. These time-bound collaborative sessions can transform dialogue into actionable knowledge. "Fishbowl" discussions or moderated roundtables can also enhance inclusivity by allowing rotating participation, ensuring that smaller states, technical experts, and civil society voices are heard alongside major actors. Digital participation platforms should complement in-person engagement, enabling remote contributions, real-time feedback, and broader global representation. Additionally, curated exchanges between policymakers and practitioners — including local government officials, public managers, and frontline implementers — would ground discussions in operational reality. Finally, the Dialogue could establish ongoing thematic communities of practice that continue collaboration beyond the event itself, supported by knowledge repositories and virtual meetings. By combining deliberative, experimental, and output-oriented formats, the AI Dialogue can move beyond symbolic engagement toward substantive international cooperation on AI governance.

Several existing policies and initiatives offer valuable foundations for effective AI governance. At the normative level, the OECD AI Principles have provided widely adopted guidance on trustworthy AI, influencing national strategies across both developed and emerging economies. Similarly, UNESCO's Recommendation on the Ethics of Artificial Intelligence represents the first global standard-setting instrument in this domain, emphasizing human rights, inclusiveness, and sustainability. In the regulatory sphere, the European Union's AI Act stands out as a comprehensive risk-based framework that categorizes AI systems according to potential harm and establishes obligations accordingly. This approach offers a structured model for balancing innovation with safety and accountability. Complementary efforts by the Council of Europe on AI and human rights further strengthen the legal dimension of governance. Operationally, several governments have developed practical tools to support implementation. For example, algorithmic impact assessment frameworks and public procurement guidelines for AI systems help public administrations evaluate risks before deployment. National digital service units and public sector innovation labs in countries such as Canada, Singapore, and the United Kingdom have also piloted governance-by-design approaches that integrate ethics, transparency, and oversight into system development processes. Multi-stakeholder platforms, including partnerships facilitated by international organizations and development banks, contribute by enabling knowledge exchange and capacity-building across regions. Open data initiatives and transparency portals further enhance accountability and public trust. Together, these policies, practices, and platforms demonstrate that effective AI governance requires a layered approach combining ethical principles, regulatory mechanisms, institutional capacity, and practical implementation tools adapted to diverse national contexts.