Vitalism Civic Innovation

A successful first Global Dialogue would produce three concrete outcomes. First, a commitment to interoperable governance frameworks rather than competing blocs. AI governance risks fragmentation along geopolitical lines. Success means agreeing on portable principles—data sovereignty, community consent, open licensing—that allow different jurisdictions to connect without forcing uniform rules. This is the "islands of interoperability" model: distinct but connected. Second, a dedicated funding stream for infrastructure-appropriate AI. Many Indigenous and rural communities lack reliable electricity, broadband, or compute. A successful Dialogue would recognise that frontier AI models are irrelevant where basic infrastructure is absent, and instead fund small, usable tools (offline digitisation, community radio integration, lightweight speech recognition) that match local conditions. Third, cultural production as a formal pillar of AI governance. Technology alone cannot keep languages alive. The Dialogue should mandate that any AI project for Indigenous or minority languages pair technical workstreams with artist, educator, or media tracks—music, storytelling, radio news, place-based reporting. Community radio models show this works. Finally, success would be measured by reusable, community-governed infrastructure, not one-off pilots. A concrete example: translating a major AI governance resource into a language spoken by millions under open licences creates lasting infrastructure. The Dialogue should commit to scaling such models across multiple language communities within two years.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Open-source software, open data and open AI models
• Transparency, accountability, and human oversight
• Interoperability of governance approaches

Interoperability of governance approaches is the most urgent priority. Rising geopolitical fragmentation threatens to turn AI governance into competing silos. A neutral, network-based diplomacy model shows that interoperability-not alignment-allows diverse systems to connect while respecting local sovereignty. This is essential for cross-border and stateless language communities where no single state represents the group. Open-source software, open data and open AI models are foundational because closed, externally controlled datasets turn AI into an engine of exclusion. Community-owned corpora, published with protected-but-permissive licences, prevent enclosure and enable reuse. Models such as community-governed open lexica and speech corpora demonstrate how local stewardship makes open approaches work. Social, economic, ethical, cultural, linguistic and technical implications captures the core insight that technology alone is insufficient. Cases across different continents show how script choices, diaspora fragmentation, and uneven data create sovereignty challenges. Culture is the multiplier-music, storytelling, place-based media keep languages socially relevant so technical investments compound rather than stagnate. Transparency, accountability, and human oversight are critical because governance failures-consent violations, opaque licensing, external control-repeatedly undermine Indigenous language AI. Community-owned data pipelines (collection → consent → maintenance → reuse) only function when local institutions can exercise oversight. Transparency about data use and accountability to community-defined protocols are non-negotiable for building trust and preventing harm.

First, the infrastructure gap as a governance category. The listed themes treat AI as primarily a software or policy problem. But in many Indigenous and rural regions, the absence of basic digital infrastructure (reliable electricity, broadband, local compute, sustainable device access) renders even well-designed governance frameworks unworkable. We need a new theme: infrastructure-appropriate AI-which recognises that scaling AI for global equity requires material investment in rural connectivity, distributed data hubs, and offline-first tools. Second, transnational community data governance for stateless languages. Languages like Punjabi, Quechua, and Gondi cross multiple nation-state jurisdictions. No single country represents them. Current governance themes assume national sovereignty as the default. We need explicit mechanisms for diaspora-led, cross-border data trusts that operate without a host state-what might be called multi-territorial language sovereignty. Third, cultural institutions as AI infrastructure. Community radio stations, local hubs, museums, and digital humanities archives are currently treated as "applications" of AI. In practice, they are essential data generators, distribution channels, and governance anchors. The missing theme is cultural infrastructure as core AI governance, which would require every technical project to be paired with an artist, educator, or media partner from the outset.

In the Indigenous and minority language preservation sector, governance gaps across the four selected thematic areas create both acute challenges and strategic opportunities. Challenge – Interoperability gap: No common standards exist for data sovereignty across borders. A language like Quechua or Punjabi spans multiple countries, each with different AI regulations. This fragmentation prevents communities from pooling corpora or building shared tools, forcing each diaspora fragment to start from zero. Challenge – Transparency deficit: Most commercial AI models trained on minority language data lack meaningful consent mechanisms. Communities rarely know what data is collected, how it is used, or who profits. The result is distrust and withdrawal—communities refuse to contribute to datasets, stalling preservation efforts. Challenge – Human oversight failure: Local institutions (community radio stations, hubs, schools) have the trust but lack the technical capacity to exercise oversight. External researchers or firms collect data, but accountability loops back to funders or universities, not to speakers themselves. Opportunity – Open infrastructure as a public good: The same gaps create demand for community-owned open lexica, speech corpora, and small-scale tools under protected-but-permissive licences. Pilot models (community hubs with participatory data pipelines) show this works when paired with capacity-building for local oversight committees. Opportunity – Interoperability as a diplomatic asset: Neutral, network-based actors can broker interoperability agreements between different national frameworks. This opens pathways for cross-border language data trusts—a governance innovation currently absent from most AI dialogues. The most significant opportunity is to move from one-off pilots to durable, community-governed ecosystems where transparency and accountability are designed in from the start, not added after harm occurs.

The AI Dialogue can advance international cooperation by shifting from competing national frameworks to interoperable governance ecosystems. Its most valuable role is as a neutral convening platform—not to force uniform rules, but to enable distinct systems to connect. First, the Dialogue should establish minimum interoperability standards for cross-border data flows in sensitive domains like Indigenous languages. This includes portable consent protocols, recognised data trust models, and mutual recognition of community-governed licences. Without these, languages spoken across multiple countries remain fragmented. Second, the Dialogue can pioneer multi-stakeholder cooperation beyond state actors. Indigenous language governance involves community radios, digital hubs, museums, universities, and diaspora networks. The Dialogue must bring these non-state actors into formal decision-making, not just as consultants but as co-designers of governance protocols. Third, the Dialogue should create a mechanism for translating governance principles into infrastructure investments. Too often, AI governance remains abstract. The Dialogue can link interoperability agreements to concrete funding for open lexica, offline tools, and community oversight bodies—turning paper commitments into usable infrastructure. Finally, the Dialogue can act as a clearing house for governance learning. Communities experimenting with data trusts, protected licences, or community radio-integrated AI need to share what works without reinventing the wheel. A living repository of governance models, legal templates, and technical standards would accelerate cooperation while respecting local autonomy. In short, the Dialogue's added value is interoperability without homogenisation—connecting islands of governance rather than flooding them.

The AI Dialogue should build upon several existing initiatives while bringing distinct added value. Existing initiatives to connect with: UNESCO International Decade for Indigenous Languages (2022–2032) and its Global Action Plan – already supports community-driven digital preservation. Creative Commons + community governance pilot models (e.g., protected-but-permissive licences with local oversight) – tested in several language communities. Community radio networks (e.g., Yacouba, Quechua, Gondi grassroots stations) – existing distribution and data-collection infrastructure. National open research forums and rural hub networks (e.g., national connected hub systems) – physical spaces for governance capacity-building. Cross-border data trust experiments for stateless languages – emerging but fragmented. What the AI Dialogue can add: First, scale through interoperability. These initiatives operate in isolation. The Dialogue can create lightweight connection protocols, allowing a Quechua data trust to share governance learnings with a Punjabi diaspora network without merging their systems. Second, a governance lens on infrastructure funding. Current initiatives often receive technology funding without governance capacity. The Dialogue can mandate that any funded project include a transparent, accountable human oversight body rooted in the community. Third, legal and technical templates. The Dialogue can produce open-source governance toolkits—consent forms, data trust charters, licence suites – adapted for low-connectivity, multilingual, cross-border contexts. Finally, political legitimacy. Many community-led initiatives lack recognition from national or international bodies. The Dialogue can formally recognise community-governed data infrastructures as legitimate governance mechanisms, unlocking access to broader cooperation frameworks.

Different stakeholders should contribute according to their distinct roles, with the Dialogue structured to enable parity rather than hierarchy. Recommendations by stakeholder type: National governments should present interoperability case studies and commit to mutual recognition of community-governed data licences. Multilateral bodies (UNESCO, ITU, UNDP) should provide existing frameworks (e.g., International Decade for Indigenous Languages) and offer technical assistance. Community-based organisations (radio stations, language nests, digital hubs) must have dedicated speaking slots and voting-like roles on working groups, not just observer status. Technical open-source communities should demonstrate small, usable tools (offline speech recognition, dictionary apps) and co-design interoperability standards. Private sector should disclose training data sources for minority languages and commit to transparency audits. Structural recommendations: Pre-Dialogue community caucuses – regional gatherings (virtual + offline) to produce shared positions before the main event, with travel support. Parallel tracks – separate streams for legal/interoperability negotiations, technical standards, and cultural infrastructure, with cross-cutting synthesis sessions. Outcome-oriented working groups – not open-ended discussion, but groups with mandated deliverables (e.g., a draft interoperable consent protocol within 12 months). Proportional time – at least 40% of speaking time allocated to community and civil society representatives. The Dialogue should avoid "panel-heavy" formats. Instead, use facilitated roundtables with designated community rapporteurs and live translation into at least 10 Indigenous languages as a demonstration of commitment.

Three categories are severely underrepresented. 1. Indigenous and minority language speakers without state representation. Languages such as Punjabi, Quechua, Gondi, and Yacouba cross multiple borders. No single government advocates for them. Their perspective—on data sovereignty without a host state, on stateless governance—is almost never heard. Inclusion mechanism: Create a Stateless Language Communities Constituency within the Dialogue, with dedicated seats and a rotating representation model based on linguistic families. Provide funding for community-selected delegates. 2. Rural and low-connectivity communities. Current AI governance discussions assume stable internet, devices, and electricity. Communities lacking these are excluded by default. Their perspective is not niche—it is the reality for most of the world's 7,000 languages. Inclusion mechanism: Establish offline-first consultation rounds using community radio networks, SMS surveys, and physical workshops in rural hubs. Synthesise findings into the Dialogue through designated rural rapporteurs. 3. Community radio and cultural practitioners. Technologists and lawyers dominate AI governance. Radio producers, storytellers, musicians, and educators—who actually keep languages alive—are rarely invited. They hold practical knowledge about what works. Inclusion mechanism: Mandate that every national delegation include at least one cultural practitioner. Reserve 30% of speaking time in technical sessions for community media representatives. Cross-cutting inclusion principle: Provide travel, interpretation, childcare, and per diems. Without material support, inclusion is rhetorical.

Beyond traditional panels and keynotes, four innovative formats would drive meaningful engagement. 1. Governance simulation exercises. Participants (governments, community representatives, tech firms) role-play a cross-border language data dispute. They must negotiate consent, licensing, and oversight in real time. This builds empathy and reveals governance gaps faster than papers. Outcomes feed directly into working groups. 2. Community radio integration. Livestream Dialogue sessions to hundreds of community radio stations. Allow real-time voice message feedback via basic phones. Compile these audio inputs into daily synthesis reports read aloud on air and brought back into plenary. This bridges the connectivity divide. 3. "Unconference" solution labs. Half-day, participant-driven sessions focused on a single, concrete problem: e.g., "Design a portable consent protocol for Quechua speech data." Small groups produce draft templates or technical specifications by the end of the lab. These become working documents, not just discussion notes. 4. Cultural production showcases paired with policy sessions. Every technical session on data or interoperability is immediately followed by a 20-minute performance, short film, or storytelling set from an Indigenous artist or radio producer. This reinforces that culture is not decoration—it is infrastructure. It also energises participants and shifts power dynamics. 5. Satellite hubs. Host simultaneous Dialogue nodes in rural locations (e.g., a connected hub in a Gaeltacht region or a community radio station in Côte d'Ivoire). Feed their discussions directly into the main stage as live breakout rooms. This distributes authority rather than centralising it. These formats prioritise action over performance, inclusion over spectacle, and culture over technical abstraction.

Several concrete examples demonstrate effective AI governance in practice, particularly for Indigenous and minority language contexts. Policy example - Protected-but-permissive licensing. Community-governed Creative Commons licences add a local oversight layer to open licensing. Data can be used for non-commercial preservation, but commercial use requires community consent. This balances openness with sovereignty and has been tested in multiple language communities. Practice example - Participatory data pipelines. A community hub model implements four stages: collection (with informed consent), consent (documented and revocable), maintenance (community stewards), and reuse (tracked and audited). This pipeline ensures transparency and accountability are designed in, not added later. Platform example - Community radio-integrated AI. In several regions, local radio stations serve as both data collectors (recording speech, stories, news) and distribution channels (broadcasting AI-powered language tools via basic phones). This works offline, reaches low-connectivity areas, and keeps governance rooted in trusted local institutions. Approach example - Translating governance resources into Indigenous languages. A major university-UNESCO MOOC on AI and digital transformation was translated into Quechua under open licence, serving 8-12 million speakers. This transforms a one-off translation into reusable language infrastructure that supports inclusive, rights-based digital governance. Interoperability mechanism - Cross-border data trusts for stateless languages. Emerging models allow diaspora communities across multiple countries to pool speech corpora under a single governance framework recognised by each jurisdiction. This prevents fragmentation without requiring national uniformity. Practice example - Small, usable tool stacks. Rather than large language models, effective governance prioritises offline digitisation workflows, automatic speech recognition for low-resource languages, transliteration tools, and dictionary apps that match local bandwidth, skills, and electricity availability. These examples shift AI governance from principle to practice, prioritising community ownership, transparency, and interoperability over enclosure and abstraction.