UN Permanent Forum on Indigenous Issues; University of Latvia Livonian Institute

A successful first Global Dialogue on AI Governance should produce binding recommendations for AI development — not merely high-level declarations, but outcomes accompanied by accessible communication products such as infographics and plain-language materials that can be understood and used across the full range of stakeholders involved. That range on stakeholders is genuinely diverse. It extends from large technology companies and international bodies all the way to small developers, local administrations, and communities on the ground. For the Dialogue's findings to reach and be actionable by this entire spectrum, the format of outcomes matters as much as their content. The critical risk to avoid is a disconnect between the Dialogue's conclusions and their actual implementation. High-level documents, if not deliberately bridged to practice, tend to remain at the level of principle — leaving the gap between global governance frameworks and real-world application intact. A successful Dialogue would therefore not only articulate binding recommendations but also ensure those recommendations are translated into concrete guidance for implementation, including for actors with limited institutional capacity. In short, success means closing the distance between what the Dialogue concludes and what stakeholders at every level — from international institutions to local communities — can actually do with those conclusions.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Safe, secure and trustworthy AI
• AI capacity-building

Safe, secure and trustworthy AI is a priority because AI is increasingly replacing search engines and traditional information sources. In this role, contextual precision and the accurate reflection of information become essential - not only within the AI ecosystem itself, but for the broader information environment it shapes. When AI systems serve as primary gateways to knowledge and decision making, the stakes of unreliable or contextually inadequate outputs extend well beyond the technology sector. AI capacity-building is urgent because AI development actively widens existing digital gaps. Building and deploying AI systems requires large amounts of data, and this requirement places small and under-resourced communities at a structural disadvantage. Without dedicated capacity-building efforts, these communities - which include the majority of the world's linguistic and cultural diversity - will be unable to remain competitive or to participate meaningfully in shaping the AI landscape that increasingly affects them. Social, economic, ethical, cultural, linguistic and technical implications of AI are of particular concern under conditions of data scarcity and critical lack of data. In these conditions, AI systems are prone to producing biases that can pollute the digital space of endangered or under-resourced languages and communities. Beyond bias, such systems risk generating misleading outputs across a range of areas, with potentially serious consequences for communities that already have limited capacity to identify, challenge, or correct those outputs.

One critical cross-cutting issue not sufficiently captured by the listed themes is the bridging of the digital divide in AI development - specifically as it affects under-resourced languages and communities, including indigenous peoples, smaller languages and communities. These groups, small in numbers or under-resourced in many ways, in fact represent the majority of the world's linguistic and cultural diversity. Yet they remain either excluded from AI development processes due to a lack of sufficient resources, or improperly addressed within them. AI approaches are, in many cases, developed based on the experience of larger or globally dominant groups, and this leaves a direct imprint on outcomes and application. The needs of under-resourced communities, the ways in which AI is used and applied in their contexts, and the specific approaches required to develop AI under conditions of critical resource scarcity are substantially different from the assumptions embedded in mainstream AI development. These differences remain largely unexplored - not only in relation to AI, but across the digital landscape as a whole. Addressing this gap is not simply a matter of equity, though it is that too. It is also a matter of the actual quality and inclusiveness of global AI governance. Frameworks developed without meaningful input from the world's most linguistically and culturally diverse communities will inevitably reflect the priorities and assumptions of a narrow subset of stakeholders. A truly global dialogue on AI governance must therefore explicitly address how AI development can be made appropriate, accessible, and genuinely useful for under-resourced languages and communities - and how those communities can be meaningfully included in shaping the norms and tools that will increasingly affect their lives.

Governance gaps and uneven AI development have a direct and serious impact on the communities I work with — critically under-resourced language and cultural communities that are already lagging behind and experiencing a rapidly expanding digital gap. The core challenge is twofold: these communities lack the resources to engage with AI development on their own terms, while those building and deploying AI solutions largely lack knowledge of their specific needs and conditions. The result is a compounding problem — under-resourcing produces invisibility, and invisibility produces further under-resourcing. This has consequences beyond the technical. It generates a biased informational space in which AI systems fail to accurately reflect these communities' languages, cultures, and realities — and as AI increasingly shapes how information is produced and trusted, this bias becomes self-reinforcing. The broader effect is exclusion: from the dialogue about how AI should be governed, and from the benefits AI could offer. Given that under-resourced language and cultural communities represent the majority of the world's linguistic and cultural diversity, their absence from AI governance processes means global frameworks are being shaped without the input of a significant portion of humanity.

The AI Dialogue is a first-of-its-kind instrument for addressing AI development and its related challenges at the international level. It has a unique opportunity to function as a genuine connecting platform — bringing together the key stakeholder groups whose engagement is essential for meaningful progress: governments, communities, academia, and industry. Each group plays a distinct and indispensable role. Governments set the regulatory and policy frameworks within which AI operates. Communities are the ultimate users and those most directly affected by AI's consequences. Academia generates the knowledge base that informs responsible development. Industry — spanning large technology companies and small developers alike — is where AI solutions actually become reality. This last point deserves particular emphasis. Whatever frameworks or recommendations the Dialogue produces, their implementation depends directly on academia and industry actors. Securing not merely their participation but their genuine interest and active contribution is therefore critical. A dialogue that fails to engage those who actually develp ideas behind, build and deploy AI will struggle to translate its conclusions into practice. The Dialogue's role in advancing international cooperation is thus not only to produce agreed frameworks, but to build the relationships and shared understanding among stakeholder groups that make sustained cooperation possible — including by ensuring that less visible stakeholders, such as under-resourced communities and small developers, have a real voice alongside the more powerful actors who typically dominate global governance processes.

Several existing initiatives offer a strong foundation for the AI Dialogue to build upon. First, UNESCO's recently launched Global Roadmap on Multilingualism in the Digital Era provides directly relevant frameworks and commitments whose implementation plan could be cross-linked with the AI Dialogue's outputs, creating coherence between multilingualism policy and AI governance. Second, the United Nations Permanent Forum on Indigenous Issues (UNPFII) is an established platform for engagement with indigenous communities worldwide. Connecting the AI Dialogue with this mechanism would ensure that indigenous peoples' perspectives are systematically integrated into AI governance discussions. Third, the First Global Survey of Indigenous Languages contains a substantive and evidence-based section directly addressing the digital situation of indigenous language communities, including AI. It reflects actual conditions on the ground and should inform the Dialogue's understanding of what inclusive AI governance means for the world's most under-resourced linguistic communities. The added value the AI Dialogue could bring is to serve as a cross-cutting connector among these and similar initiatives — ensuring that their findings and priorities are brought into direct dialogue with the actors shaping AI development and governance. Many relevant frameworks and bodies already exist but operate in relative isolation from the AI governance space. The Dialogue is well positioned to close that gap.

Participation, providing information and involvement in the discussions, including co-creating of outcomes.

Indigenous people, Digitally underpowered languages and communities

Citizen science platforms or similar engagement platforms for diverse landscape of stakeholders, to avoid limited participation and ensure regular work on and elaboration of outputs.

Several planned COST actions in Europe