ONG AMARANTA

It would ensure that civil society and its experiences are considered central, participating alongside decision-makers and contributing their grassroots and territorial perspectives on the various facets of the discussion surrounding artificial intelligence: social, labor, innovation, inclusion, gender, and human rights. The discussion must be multi-stakeholder and consider intersectionality. A resounding success would involve participation from all areas of development and industry, leading to a mutually agreed-upon declaration that enables sustained work in research and education on these topics.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Safe, secure and trustworthy AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

his co-governance model requires institutionalizing the participation of grassroots organizations, social movements, academia, and the private sector in permanent working groups. The protection and promotion of human rights must operate as the binding legal framework that guides the entire process, with special attention to intersectionality (gender, ethnicity, disability, and territory) so that differentiated impacts are made visible and addressed. Transparency means opening algorithmic black boxes through accessible citizen audits and language that is understandable to non-technical communities. Accountability requires establishing clear responsibility mechanisms: companies must answer for biases and externalities; states, for guaranteeing rights; civil society, for independent monitoring. All of this must be underpinned by effective human oversight, ensuring that no automated decision strips away human agency or dignity. The success of this approach materializes in a mutually agreed-upon declaration containing binding commitments and a roadmap for sustained work in research and education. Only with broad, pluralistic participation and robust safeguards can we develop AI that is not merely innovative, but genuinely trustworthy, fair, and in service to all people.

While the listed themes (human rights, transparency, accountability, human oversight, and the multidimensional implications of AI)are essential pillars, there are cross-cutting and emerging issues that deserve explicit incorporation into a comprehensive governance framework. These dimensions do not fit neatly into existing categories yet are critical for ensuring that AI governance is truly holistic and responsive to the complexities of the present moment. Ecological sustainability and environmental impact represent one such cross-cutting issue. The AI lifecycle: from energy-intensive model training and data center water consumption to the extraction of critical minerals and electronic waste disposal, carries a significant environmental footprint that disproportionately affects vulnerable communities. This intersects directly with environmental justice, human rights, and intergenerational equity. A governance model centered on civil society must embed ecological criteria as a binding constraint, not an afterthought. Similarly, epistemic justice and cognitive sovereignty address how AI systems shape what knowledge is validated and whose voices are represented. Indigenous peoples, Afro-descendant communities, and linguistic minorities often face exclusion or distortion by dominant AI models. Ensuring cultural and linguistic diversity requires intentional design choices that go beyond mere translation to protect epistemic autonomy. Finally, technological sovereignty and the democratization of infrastructure raise fundamental questions about who controls the means of AI production. The concentration of development in a handful of corporations and nations creates dependencies that constrain local agency and democratic participation. A robust governance framework must address open infrastructures, public alternatives, and community-owned data governance models. Incorporating these emerging issues does not dilute the foundational pillars but rather deepens them, ensuring that AI governance is ecologically sustainable, culturally plural, and genuinely democratic.

In my region (Latin America) and across the sectors I engage with (particularly civil society organizations, grassroots movements, and public interest technology) the governance gaps in the thematic areas identified above manifest acutely. Ecological sustainability remains a critical blind spot. Countries in the Global South are often positioned as extractive frontiers for AI infrastructure: sites for mineral mining (lithium, copper, rare earths) and e-waste dumping, yet excluded from decisions about AI governance. The governance gap here is the absence of binding environmental impact assessments for AI supply chains, leaving communities vulnerable to dispossession without recourse. Epistemic justice faces a deepening gap. Most AI models are trained on datasets that marginalize Spanish, Portuguese, and especially Indigenous languages. In my region, this threatens not only cultural survival but also access to justice and public services. The opportunity lies in fostering regional coalitions for open, plurilingual foundational models developed with community participation—but this requires sustained investment and political will that current governance frameworks do not prioritize. Technological sovereignty is perhaps the most pronounced gap. Dependence on foreign corporations for cloud infrastructure, AI platforms, and even data storage constrains the ability of governments and civil society to exercise meaningful oversight. The challenge is structural: concentrated ownership limits accountability mechanisms. Yet the opportunity is emerging through South-South cooperation, public digital infrastructure initiatives, and grassroots technology cooperatives that are experimenting with alternative, community-governed AI models. The most significant challenge is that governance gaps are often treated as technical problems rather than structural ones. The opportunity lies in building multi-stakeholder, territorial coalitions capable of translating local experiences into binding regulatory frameworks, transforming these gaps into sites of democratic innovation.

The AI Dialogue can serve as a critical catalyst for international cooperation by creating a structured, inclusive space where diverse stakeholders (governments, civil society, academia, and industry) can move beyond fragmented national approaches toward shared principles and coordinated action. First, the Dialogue can help bridge the North-South divide in AI governance. Currently, regulatory frameworks are being shaped predominantly by a handful of Global North jurisdictions, often without adequate representation of the priorities and vulnerabilities of countries in the Global South. The Dialogue can ensure that perspectives from regions facing extractive supply chains, linguistic exclusion, and infrastructure asymmetries are not merely consulted but actively shape the agenda. This is essential for preventing governance from becoming a new axis of geopolitical inequality. Second, the Dialogue can advance convergence on binding principles without imposing homogeneity. By facilitating sustained deliberation across cultural and legal traditions, it can help identify core non-negotiables such as human rights protections, environmental sustainability, and meaningful human oversight while allowing flexibility for contextual implementation. This balance between common standards and local adaptation is crucial for legitimacy and effectiveness. Third, the Dialogue can establish mechanisms for mutual accountability and shared learning. International cooperation often fails at implementation. The Dialogue can foster commitments to transparency, shared auditing protocols, and cooperative enforcement mechanisms. It can also serve as a platform for pooling resources toward public-interest AI research, open infrastructure, and capacity-building (particularly for historically marginalized communities). Ultimately, the AI Dialogue's most valuable role is to institutionalize the principle that AI governance is a global public good. In doing so, it can transform competition into collaboration, ensuring that the development of artificial intelligence is guided not by the interests of the few but by the rights and aspirations of all.

The AI Dialogue could bring distinct added value by serving as an integrative bridge across these existing mechanisms. While UNESCO focuses on ethical standards, GPAI on multistakeholder expertise, and the Council of Europe on legally binding frameworks, the AI Dialogue could offer a flexible, agile space for inclusive dialogue that connects these silos. It could prioritize civil society and grassroots participation more systematically than existing platforms, ensuring that territorial and intersectional perspectives inform global governance. Additionally, the Dialogue could focus on implementation and capacity-building, particularly for underrepresented regions in the Global South, transforming fragmented initiatives into a coherent ecosystem for collaborative action.

Civil society organizations and grassroots movements should contribute their territorial knowledge and lived experiences, ensuring that governance frameworks reflect the realities of marginalized communities. They can serve as watchdogs, monitoring accountability, and as bridges connecting global discussions to local implementation. Governments can provide political legitimacy and the authority to translate dialogue outcomes into binding regulations. They should commit to implementing agreed principles domestically and allocate resources for capacity-building, particularly in underrepresented regions. Academia and research institutions can offer evidence-based analysis, contribute to impact assessments, and develop methodologies for auditing and evaluation. They can also serve as neutral conveners and institutional memory. Industry and technology developers should bring technical expertise and transparency about system capabilities and limitations. Their contribution lies in committing to responsible innovation practices, participating in co-design processes, and being accountable for externalities. Multilateral organizations and UN agencies can provide platforms, technical assistance, and mechanisms for scaling successful initiatives across borders, ensuring coherence with existing frameworks.

Despite growing multistakeholder rhetoric, several voices and communities remain systematically underrepresented in global AI governance discussions. Their exclusion perpetuates governance frameworks that are ill-equipped to address diverse realities. Indigenous peoples and communities are frequently absent, despite bearing the impacts of extractive supply chains and facing cultural erasure through datasets that marginalize their languages and knowledge systems. Their cosmovisions—often emphasizing relationality, reciprocity, and ecological stewardship—could fundamentally enrich ethical frameworks. Women, particularly in the Global South, remain underrepresented in technical and decision-making spaces. Gender-based violence facilitated by AI, algorithmic bias in employment, and the care economy's transformation receive insufficient attention without their sustained participation. Rural and peri-urban communities are often invisible, as discussions center on urban, high-connectivity contexts. Yet these communities experience AI-driven agricultural technologies, infrastructure decisions, and service delivery in ways that differ markedly from metropolitan experiences. Workers and trade unions in the informal economy, gig economy, and traditional sectors have limited seats at governance tables. Their knowledge of labor precarization, algorithmic management, and workplace surveillance is irreplaceable. Youth and future generations are frequently tokenized rather than empowered as co-decision-makers, despite being most affected by long-term AI trajectories. Persons with disabilities bring critical insights into accessibility, algorithmic discrimination in healthcare and employment, and the design of assistive technologies—yet are often consulted only as an afterthought. Linguistic minorities face exclusion when discussions default to English and when AI models ignore non-dominant languages.

Rather than top-down knowledge transfer, peer-to-peer learning exchanges connect communities, grassroots organizations, and practitioners across regions facing similar challenges. For example, a network of Indigenous communities grappling with data sovereignty could exchange strategies, while trade unions from different countries could share experiences with algorithmic management. These exchanges build solidarity, amplify practical knowledge, and reduce dependency on expert-driven frameworks. Collaborative sandboxes provide safe, time-bound spaces where stakeholders co-design and test governance approaches in real or simulated environments. Unlike regulatory sandboxes driven primarily by industry, collaborative sandboxes include civil society oversight, community feedback loops, and iterative refinement. Participants could prototype community data trusts, public algorithm registries, or participatory auditing mechanisms—turning dialogue into action.