Olabi.Org

In my view, a successful first Global Dialogue would establish AI governance as genuinely multilateral. Success means visible influence from Latin America, Africa, the Caribbean, the Pacific and Asia on the agenda, the Scientific Panel's evidence base, and any outcome documents (not as observers ratifying frameworks designed elsewhere, but as co-chairs and panel speakers shaping them). From my work in Brazil, I see how often Global South perspectives are invited into AI conversations late, after key framings are already set. The Dialogue also would expand the scope of AI governance beyond content and outcomes to include system design itself. Recommendation systems, engagement architectures, interface choices and optimization targets are not neutral — they shape behaviour, information flows and democratic participation, and they amplify or mitigate inequalities. A successful first session would put these design layers explicitly on the governance agenda. It would also recognise designers, developers and AI practitioners as co-producers of governance, not only as subjects of regulation. Without this bridge between policy and technical communities, governance frameworks risk being unimplementable. And the Dialogue would adopt an intersectional understanding of who is harmed by AI. Risks are not distributed equally; women, racialised populations, older adults, LGBTQIA+ people, peripheral populations and people with disabilities experience them differently and cumulatively. Success means a Dialogue that is multilateral in fact, governs design and not only outputs, and takes intersectionality seriously.

• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Transparency, accountability, and human oversight
• Protection and promotion of human rights

These four priorities reflect where my work intersects with the most urgent gaps in current AI governance. AI capacity-building is foundational. As Co-Executive Director of Olabi, I lead programs that have trained thousands of people from groups historically underrepresented in technology - Black women, older adults, peripheral communities, Black professionals entering the tech workforce. From this experience, I have learned that capacity-building must be intersectional and audience-specific, channeled through organizations that already serve excluded communities, and designed to bring people in not only as users of AI but as critical participants in shaping it. Without this, capacity-building will reproduce the very asymmetries it claims to address. Protection and promotion of human rights is where AI's distributional injustices become most visible. Transparency, accountability and human oversight should not stop at content-level moderation. They must reach the design layers that shape AI behavior: recommendation systems, engagement architectures, interface choices, optimization targets and data pipelines. These choices are not neutral, and they remain largely outside regulatory scrutiny, particularly in the Global South. This is where I see the most urgent work to be done. Social, economic, ethical, cultural, linguistic and technical implications matter because Global South realities are still treated as edge cases in AI development. Portuguese-language gaps in foundation models, age-based exclusion, peripheral and Indigenous contexts, these implications shape whether AI strengthens or weakens democratic life worldwide.

Three cross-cutting issues deserve more explicit attention in the Dialogue. System design as a site of governance. Current AI governance frameworks focus largely on content and outcomes (what AI produces, what it decides, what it disseminates). They give comparatively little attention to the design choices that shape those outcomes: recommendation systems, engagement architectures, interface decisions, the data pipelines that feed AI systems and the optimization targets that drive them. These choices are not neutral. They structure attention, influence which information circulates, and amplify or mitigate existing inequalities. They remain largely absent from regulatory approaches, particularly in the Global South. The Dialogue should treat design as governance, not only as engineering. Technical communities as co-producers of governance, not only as subjects of regulation. AI governance debates tend to take place between policymakers and civil society, with the designers, developers and AI practitioners who actually build these systems engaged only as objects of regulation. The result is policy that is hard to implement and engineering practice that is structurally insulated from public-interest concerns. The Dialogue should support methodologies that bring these three communities into structured, sustained dialogue - a practice my work in Brazil has tested in the context of platform regulation and online gender-based violence. Of all axes of digital and AI exclusion, age is the most overlooked globally. With populations ageing rapidly - particularly in the Global South - governance frameworks that ignore older adults will systematically fail a growing share of humanity. In Brazil, 42% of people aged 60 or older have never used the internet.

Brazil and Latin America offer a particularly instructive case for this Dialogue, combining significant capacity with deep governance gaps. Brazil is the largest digital market in Latin America, yet AI deployment here is reproducing structural inequalities at scale. Capacity-building remains fragmented and concentrated in urban centers and elite institutions, leaving Black women, older adults, peripheral and rural populations, Indigenous communities and LGBTQIA+ people systematically underserved. In Brazil, 42% of people aged 60 or older have never used the internet — a striking marker of how ageing intersects with AI exclusion. Human rights frameworks struggle to keep up with platform-mediated harms; transparency and accountability obligations stop at content moderation and rarely reach the design layers that actually shape AI behavior. Civil society organizations working with affected communities are often the most knowledgeable actors on these harms, yet the least resourced. Brazil presided over the G20 in 2024 with a digital and AI agenda focused on inclusion. Its Supreme Court has recently articulated a duty of care principle for digital platforms — a development that opens space for design-centered governance. Brazilian civil society has built strong methodologies for bridging policy, technical and community perspectives, including through high-level multilateral spaces such as the December 2025 MESECVI Conference of the OAS in Fortaleza. The Dialogue can amplify these approaches and channel resources to organizations already serving historically excluded communities — converting Brazil's experience into shared Global South practice.

The AI Dialogue can correct the geographical asymmetry of AI governance. Most existing AI governance arenas are convened by, or originate in, the Global North. To realize this, the Dialogue must reserve thematic co-chair roles, panel speakers and Scientific Panel evidence inputs for Global South actors — civil society included. It also can connect AI governance to the United Nations human rights system. Cooperation on AI cannot be reduced to technical interoperability between national rules. It must be grounded in international human rights law. And the Dialogue can build bridges between policy, civil society and technical communities as international cooperation on AI is too often a conversation between governments and large technology companies. The Dialogue should formally include the designers, developers and AI practitioners who build these systems (not as observers, but as co-producers of governance frameworks) alongside civil society organizations working with affected communities. Without this triangulation, governance frameworks will remain technically unimplementable and engineering practice will remain insulated from public-interest concerns. The Dialogue's most distinctive contribution would be to institutionalize that triangulation.

The UN human rights mechanisms; multilateral processes on digital governance and civil society networks. The Dialogue's most distinctive contribution is being the only universal, multistakeholder UN forum where governments, civil society, technical communities, academia and affected communities meet annually to align on AI governance. This convening role is its main asset — and its main responsibility.

The Dialogue's format should reflect the diversity of contributions different stakeholders can make. Member States bring legitimacy and the capacity to translate outcomes into national policy. Civil society organizations working with affected communities bring evidence on AI harms that academic and industry sources rarely capture. Technical communities — designers, developers, AI practitioners — bring implementability: they know what works in code and what does not. Academia brings rigor and comparative analysis. Industry brings scale and resources. Recommendations for format: - reserve thematic co-chair roles for Global South civil society and technical communities, alongside Member States, in each thematic discussion. This is the single most consequential format choice for inclusivity. -create structured space for cross-community dialogue, not only sectoral panels. Sessions that bring policymakers, civil society and technical practitioners into the same conversation — with shared questions, not separate plenaries — produce outcomes that none of the three communities can produce alone. -introduce a permanent civil society constituency with capacity to coordinate inputs, nominate participants and provide continuity between sessions, drawing on lessons from the Internet Governance Forum. -make written and asynchronous contributions count. Many civil society organizations from the Global South cannot afford to travel annually to Geneva or New York. Structured written submissions, regional consultations between sessions, and remote interventions should be treated as substantive inputs to the Scientific Panel and to outcome documents — not as second-tier participation. -publish, in advance and in plain language, how stakeholder inputs will inform thematic discussions. Transparency builds trust and engagement.

Older adults: of all axes of digital and AI exclusion, ageing is the most overlooked globally. With populations ageing rapidly — particularly in the Global South — governance frameworks designed without older adults will fail a growing share of humanity. Inclusion requires intergenerational consultations, dedicated capacity-building programs, and recognition of ageism as an AI governance issue. Black women and racialized communities: disproportionately exposed to harmful AI deployments, yet rarely present where AI policy is shaped. Inclusion requires channeling resources through Black-led organizations and reserving co-chair roles in human rights and accountability discussions. LGBTQIA+ communities: affected by content moderation bias, deepfake-enabled harassment and identity-based discrimination by AI systems. Peripheral, rural and Indigenous communities: often described in AI policy as "users" or "data subjects" rather than as actors with knowledge to contribute. Inclusion requires partnerships with grassroots organizations, support for Indigenous-led data governance initiatives, and explicit attention to languages — Portuguese, Spanish, Indigenous and African languages — in foundation models, benchmarks and open data initiatives. Technical practitioners working in the public interest: designers, developers and AI practitioners outside Big Tech — those working in civil society, public institutions and small organizations in the Global South — are central to implementing any governance framework, yet are largely absent from intergovernmental forums. How to include them. Through reserved co-chair seats, multi-year participation support targeted at women-led, Black-led, intergenerational and community-based organizations, regional consultations between sessions, and explicit recognition of these communities.

Not only sectoral panels: most multilateral forums separate stakeholders into parallel tracks — governments here, civil society there, technical communities elsewhere. The Dialogue's can organize sessions that bring policymakers, civil society representatives and technical practitioners (designers, developers, AI practitioners) into the same room, around shared questions, with structured facilitation. Lived-experience panels: ledicated sessions in which people directly affected by AI systems — Black women, older adults, peripheral residents, LGBTQIA+ people, workers in AI-mediated industries — speak alongside experts, not as testimonial color but as primary evidence on AI harms. "Design walkthroughs" on real AI systems: working sessions in which technical practitioners walk policymakers and civil society through the actual design choices of AI products and platforms — recommendation systems, engagement architectures, optimization targets, data pipelines. This format demystifies governance objects that are usually discussed in abstract terms and grounds policy proposals in implementable detail.

The Digital Statute of Children and Adolescents (ECA Digital, Brazil, 2025): sanctioned in September 2025 and in force since March 2026, Law 15.211/2025 establishes preventive obligations for digital products and services likely to be accessed by children and adolescents, and is one of the first national frameworks to regulate platform design directly - including restrictions on infinite scroll, autoplay and engagement-driven features that capture attention through compulsive design. The law shifts the regulatory object from content to system architecture itself, treating design choices as governance objects. It offers a transferable model relevant to AI governance: a legal standard that reaches design and operation, not only outputs. The duty of care principle for digital platforms (Brazil, 2025): in June 2025, the Brazilian Supreme Court articulated a duty of care obligation for digital platforms, requiring diligent and preventive action - not only reactive moderation. This principle is now being extended in Brazilian policy debate to AI governance, with explicit attention to gender-based and intersectional harms. Civil society methodologies bridging policy, technical and community communities. Olabi's program "Online Misogyny: Paths Forward," developed with the Ministry of Women, the Secretariat of Social Communication of the Presidency and the United Kingdom Government, has tested structured dialogue between policymakers, technical practitioners and affected communities - a methodology adaptable to AI governance. Intergenerational AI inclusion (Transborda 60+) and Black women in technology (PretaLab, since 2017): concrete civil society models for AI capacity-building across underexamined axes of exclusion.