Data Privacy Brasil

A successful UN Global Dialogue on AI Governance (UNGD) should be measured not only by the quality of its discussions, but by its ability to generate tangible, actionable outcomes that can shape global and national governance processes. A critical outcome would be agreed definitions of "AI red lines": clear boundaries identifying applications and practices deemed unacceptable due to their risks to human rights, democracy, and societal well-being, serving as a foundational reference for regulatory frameworks and international cooperation. The UNGD should go further and advance a shared understanding of how to operationalize moratoria on AI systems that pose unacceptable risks, building on existing UN human rights guidance, including statements from the OHCHR (A/HRC/48/31, 2021) and UN Special Rapporteurs. Transparency throughout the process is equally essential to have a successful Global Dialogue. To have clarity on how non-state stakeholders are selected, and openness regarding which initiatives and case studies are elevated. This must be accompanied by a strong commitment to meaningful multistakeholder participation, while carefully distinguishing civil society organizations from the broader "non-state stakeholder" category, as these actors do not hold equivalent power, access, or influence in governance spaces. At last, the UNGD should identify clear pathways for coordination, alignment, and coherence across different governance structures and processes. This includes advancing outcomes across key governance processes and forums such as the AI Summits, WSIS, GPAI, IGF, and the CSTD Data Governance Working Group. Finally, the UNGD should identify ways to advance effective financing and capacity-building. It could build on proposals already underway within the UN, such as the Global Fund on AI, while supporting tracking mechanisms for financing and a roadmap for capacity-building initiatives to be implemented at the national level.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Interoperability of governance approaches
• Protection and promotion of human rights

We selected these four priorities because they reflect the areas where urgent international engagement is most needed to ensure that AI governance is rights-based, inclusive, and responsive to global inequalities. Safe, secure and trustworthy AI is a priority because, for the Global South, trustworthiness cannot be reduced to technical robustness alone; it must also include democratic oversight, public accountability, and protection against harmful or high-risk deployments. Social, economic, ethical, cultural, linguistic and technical implications of AI are equally central, as AI systems are already reshaping labor, access to services, cultural production, language inclusion, and power relations across societies. These impacts are unevenly distributed and often deepen existing inequalities, particularly in contexts marked by structural asymmetries, limited institutional capacity, and linguistic marginalization. Interoperability of governance approaches is critical because duplication and fragmentation across global, ad hoc, and domestic AI governance processes can exclude less-resourced actors, especially from the Global South. The UNGD can help build interoperability by identifying shared objectives and outcomes across governance processes, providing procedural guidance for multistakeholder participation, and tracking progress against commitments and deliverables. This can also support countries in aligning domestic governance approaches. Finally, protection and promotion of human rights must remain at the center of the Dialogue. Human rights provide the clearest normative foundation for identifying AI red lines, that is, applications and practices that are unacceptable due to their risks to human rights, democracy, and societal well-being. Centering human rights is also essential to addressing harms, strengthening accountability, and ensuring that governance frameworks protect people and communities rather than simply enabling market expansion.

Yes. Several cross-cutting and emerging issues should be more explicitly reflected in the Dialogue. First, environmental and sustainability issues should also be addressed in the Global Dialogue, especially in line with the likely guidance of the Scientific Panel, as well as in alignment with UNGA Resolution A/78/L.49 and the Sustainable Development Goals. Having an outcome document with policy actions coordinated with environmental commitments can be an important step. For example, the Escazú Agreement, the first environmental treaty of Latin America and the Caribbean, which promotes the rights of access to information, participation, and justice in environmental matters, and can therefore be applied to AI-related investments such as the construction of data centers. Second, the impacts of AI on education, particularly on children and youth. While LLMs and generative AI offer opportunities to accelerate research and broaden access to knowledge, they also raise concerns about cognitive and cultural biases, erosion of critical thinking, and shifts in pedagogical paradigms. For younger and more vulnerable populations, these systems also pose heightened risks to safety, well-being, and mental health. Third, the impact of AI on labour markets is another critical cross-cutting issue, in line with recent UN and ILO discussions. AI is already reshaping employment patterns, job quality, and working conditions, raising urgent questions around worker protection, reskilling, and algorithmic management and transparency. Finally, the differentiated impacts of AI across gender must be explicitly addressed. AI systems can reproduce and amplify existing biases and inequalities, as well as technology-facilitated gender-based violence (TFGBV). At the same time, women and gender-diverse people remain significantly underrepresented in the design and governance of AI systems. Addressing these gaps is essential to ensure that AI governance frameworks are inclusive, equitable, and rights-based.

Governance gaps facilitate that AI models are deployed without clear redlines and guarantees to human rights, even with the risk of "ethics dumping" - using experimental AI in countries where no regulations exist. These challenges are particularly acute in the Global South. Many countries face structural constraints in developing and enforcing robust AI governance systems and largely use systems developed elsewhere. The current AI ecosystem is highly concentrated, with most models, infrastructures, and standards being designed in a small number of countries and companies, essentially from the Global North. This creates significant asymmetries in power, representation, and accountability. Communities in the Global South are often subject to AI systems that do not reflect their linguistic, cultural, or social realities, reinforcing existing inequalities and limiting their ability to shape how these technologies are designed and used. The lack of meaningful mechanisms to hold companies accountable across jurisdictions further exacerbates these risks, allowing private, as well as public actors, to operate with limited oversight and to externalize harms. At the cultural and linguistic level, without adequate representation of local communities and indigenous cultures, LLMs risk perpetuating colonial perspectives, reproducing stereotypes, and folklorizing those communities rather than recognizing them as legitimate sources of knowledge. Finally, the lack of interoperability standards produces "vendor lock-in" dynamics, in which less affluent countries forfeit their technological sovereignty and are relegated to the role of resource and raw material providers, unable to climb the AI value chain or capture its benefits.

The UN AI Global Dialogue can play a pivotal role in advancing international cooperation by fostering minimum common understandings and shared baselines for AI governance across diverse national and regional contexts. These shared understandings should not aim at uniformity, but rather at enabling interoperability between different governance models and rights-based approaches. Countries face distinct social, economic, and institutional realities, and AI governance frameworks must remain responsive to these differences. In this sense, the Dialogue can support a model of cooperation that is both context-sensitive and globally coherent, allowing for policy diversity while promoting alignment around fundamental safeguards, principles, and objectives. The Dialogue can also help democratize a conversation that, despite its critical importance for the future of humanity, has so far been shaped largely by competition between states and corporations, as well as by broader power imbalances, technological disparities, and persistent forms of exclusion. It offers an opportunity to foster a more open, horizontal, and inclusive discussion that brings together different sectors, countries, and perspectives on more equal terms. By doing so, the UNGD can contribute not only to greater coordination across governance processes, but also to a more legitimate and representative international debate on AI governance, one that reflects the experiences, priorities, and needs of countries and communities that have historically been underrepresented in global technology policymaking.

The UN AI Global Dialogue should build upon existing multilateral frameworks and processes, particularly the Global Digital Compact (GDC), which already provides an important political foundation for digital governance within the UN system. In this context, special attention should be given to the Data Governance Working Group established under the GDC, since data governance is a foundational layer for effective and equitable AI governance. Strengthening this connection is essential to ensure that AI development and deployment are grounded in principles of social justice, inclusion, and fairness. The work of the Data Governance Working Group offers a valuable entry point for advancing discussions on data stewardship, access, interoperability, and the mitigation of structural inequalities embedded in data ecosystems. The Dialogue should also draw lessons from other global governance processes, including climate governance. The climate regime demonstrates the value of a polycentric approach that enables multilevel action by states, local governments, international organizations, and other actors, while allowing responses to be adapted to different responsibilities and contexts. It also illustrates the importance of a credible scientific body, such as the Intergovernmental Panel on Climate Change (IPCC), in informing decentralized action and supporting shared understandings across governance levels. The added value of the AI Dialogue lies in its ability to connect these existing efforts, reduce fragmentation, and promote greater coherence across institutions and processes. It can help bridge discussions on AI governance with broader debates on data governance, sustainable development, and global inequality, while creating a more inclusive space for cooperation across regions and stakeholder groups. In doing so, it can strengthen coordination, support policy alignment, and foster more equitable and context-sensitive approaches to AI governance.

Different stakeholders can contribute meaningfully to the AI Dialogue only if the process is designed from the outset to ensure inclusive, continuous, and effective multistakeholder participation. In particular, civil society participation should not be treated as peripheral, but as a necessary procedural condition for legitimacy, accountability, and rights-based AI governance. This requires attention to civil society engagement across the design, deliberation, implementation, and follow-up phases of the Dialogue. According to the draft note on themes and structure, "relevant stakeholders" will participate in the high-level governmental plenary segment, as well as in other moments, but there needs to be transparency regarding how these stakeholders will be selected (what does "relevant" mean?). Also, according to the proposed structure, there is no need for a summary of the first day, given the limited duration of the event. This could instead be incorporated into a final summary, together with a proposal for continuing the dialogue, which could take place during the IGF in December. It is crucial that the multistakeholder segments remain part of the program, especially at the end of the second day, to support the conclusion of the event. In practical terms, the Dialogue should adopt formats and structures that protect and promote pluri-participation. This includes building on existing UN multistakeholder principles and practices, appointing a civil society liaison to support the Co-Chairs before, during, and after the Dialogue, and structuring sessions in ways that enable real-time exchange among governments, civil society, the private sector, academia, and affected communities.

Global discussions on AI governance continue to underrepresent perspectives from the Global South, not only in formal decision-making spaces but across the broader AI value chain. Countries and communities in the Global South are often positioned primarily as sources of data extraction, raw materials, energy, water, land, and precarized labor, while remaining largely excluded from agenda-setting, standard-setting, and the development of domestic technological capacities. This structural imbalance limits the ability of AI governance frameworks to reflect diverse social, economic, cultural, and linguistic realities, and risks reproducing existing global inequalities. The Dialogue should also move beyond a narrow focus on "AI safety," as seen in past AI summits, and explicitly engage with existing and ongoing harms already experienced in many parts of the Global South. These include labor exploitation in data work, lack of decent work protections, unequal access to education and digital infrastructure, and threats to safety and security in real-world deployments of AI systems. AI governance must be grounded in lived realities, rather than shaped primarily by hypothetical future risks. Underrepresented voices also include affected communities, workers across AI supply chains, Indigenous peoples, local language communities, grassroots civil society organizations, and researchers from low- and middle-income countries. Their inclusion requires more than formal invitations. It depends on meaningful participation measures, including transparent selection processes, financial and logistical support, multilingual access, low-bandwidth participation options, visa support, and agenda-setting mechanisms that allow these actors to shape priorities and outcomes from the outset. The process should include a dedicated online platform with centralized and regularly updated information, background documents, and guiding questions shared sufficiently in advance, transparent opportunities for written feedback, and materials made available in the widest possible range of languages. Without this, inclusion risks becoming symbolic rather than substantive.

First, the Dialogue should establish mechanisms to engage with regional consultations and existing convenings, including spaces such as the Forum on Internet Freedom in Africa, Digital Rights in Asia-Pacific, and RightsCon. It should also connect with the Internet Governance Forum and build on its established multistakeholder practices and processes, for instance, the National and Regional IGFs. Linking the Dialogue to these existing spaces would help broaden participation, incorporate region-specific perspectives, and reduce the concentration of discussions in a limited number of global forums. Second, the Dialogue should include multistakeholder thematic working groups organized around specific policy issues. Meaningful engagement requires concrete opportunities to influence both deliberations and outputs. These working groups could bring together governments, civil society, academia, technical experts, and affected communities to produce analytical papers, recommendations, or reports that directly inform the Dialogue and its outcome documents. Finally, the formal outcomes of the Dialogue should be shaped through multistakeholder review before publication and adoption. This would help ensure that final documents reflect a wider range of perspectives and are not produced through closed or purely intergovernmental processes. Structured opportunities for written feedback, consultation on draft texts, and transparent incorporation of comments would strengthen both legitimacy and accountability. The Dialogue should also incorporate strong inclusion and follow-up mechanisms. A multistakeholder Steering Committee, combined with multistakeholder policy clusters, could help ensure diversity, balanced representation, and continuity across the process. A multi-tiered structure with regionally representative and multistakeholder-informed agenda-setting would strengthen the ability of historically underrepresented groups, particularly actors from the Global South, civil society, and academia, to shape outcomes in a meaningful way. Taken together, these formats would help make the Dialogue more dynamic, regionally grounded, and capable of producing outcomes that are informed by those most affected by AI governance decisions.

Effective AI governance depends not only on substantive rules, but also on governance practices that make decision-making more inclusive, transparent, and accountable. In this regard, the UNGD should build on existing multistakeholder approaches already recognized across the UN and related processes, including the WSIS+20 process, the Internet Governance Forum, and the NETmundial+10 Multistakeholder Guidelines. These offer concrete procedural practices for improving coordination, information-sharing, and deliberation across diverse stakeholder groups. A particularly important practice is the use of a dedicated and regularly updated online platform to centralize information, publish background materials in advance, enable transparent written feedback, and support continuity before and after in-person meetings. This should be combined with regional and intersessional consultations, including engagement on the margins of existing events and processes, so that the Dialogue is informed by regionally grounded experiences and not only by discussions in a single global venue. Another concrete approach is to adopt institutional mechanisms for meaningful participation, such as appointing a civil society liaison, structuring sessions to allow real-time exchange among governments and non-governmental stakeholders, and ensuring that outcomes are accompanied by synthesis reports explaining how contributions were considered. These practices can strengthen legitimacy and make participation more consequential. Finally, effective governance requires equitable access to participation. Financial support, low-bandwidth access, multilingual materials, visa assistance, and accessibility and safety measures are not ancillary issues, but core governance enablers. The added value of the UNGD would be to connect and adapt these existing practices specifically to AI governance, helping reduce fragmentation while making global cooperation more inclusive and implementation-oriented.