Avant.Dev

The inaugural Dialogue will succeed if it moves beyond principle declaration toward three concrete, measurable outcomes. First, a mandate for a Structural Impact Assessment for Cultural Economies in Developing Countries, delivered to the General Assembly alongside the Dialogue's substantive recommendations. The cultural and creative sectors in developing economies are simultaneously among the most affected by AI deployment and the least represented in governance design. Paragraph 39 of A/RES/80/173 recognizes this reality — the Dialogue should operationalize it. Second, a clear reporting mechanism from the Inter-Agency Working Group on AI — mandated by Paragraph 86 of A/RES/80/173 — that integrates evidence from non-state practitioners, not only Member States and multilateral institutions. AI governance is shaped as much by infrastructure concentration, compliance cost asymmetries, and capacity-building design choices as by formal regulatory frameworks. The Inter-Agency WG should be mandated to track these dimensions. Third, a commitment that Geneva 2026 is the beginning of a differentiated engagement process — not a single consultation — where SME cultural producers and creative sector organizations from developing countries are systematically included in implementation

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• AI capacity-building
• Interoperability of governance approaches
• Open-source software, open data and open AI models

As an art and technology organization deploying AI in cultural production in Mexico City, these four areas reflect the structural dimensions that most directly affect practitioners in developing economies. Social, economic, cultural and linguistic implications: AI governance cannot be evaluated without understanding its differential impact across cultural economies. In developing countries, AI infrastructure dependency is a structural condition produced by USD-denominated compute concentration and training datasets skewed toward high-resource languages and cultural contexts. Paragraph 39 of A/RES/80/173 recognizes this; the Dialogue must operationalize it. AI capacity-building: Current frameworks assume institutional capacity that does not exist in SME cultural producers across developing economies. Capacity-building that transfers standards without adapting for structural context reproduces dependency rather than reducing it. The ITU AI Readiness Framework 2.0 (Dimension 4) provides a contextualization methodology - it should be the baseline, not an optional supplement. Interoperability of governance approaches: Compliance architectures designed in advanced economies impose asymmetric costs in developing country contexts. An interoperability framework must address not only technical standards but regulatory burden calibration - SMEs in developing countries need differentiated compliance pathways. Open-source and open data: Local data infrastructure is a prerequisite for culturally relevant AI. Open models trained on locally curated cultural datasets represent the only viable path to AI that reflects the diversity of human expression. This is a cultural sovereignty argument, not only an efficiency one.

Yes. The most significant cross-cutting issue not captured by the listed themes is the structural asymmetry between AI governance frameworks designed in advanced economies and the capacity of developing country practitioners - particularly SME cultural producers - to comply with, benefit from, or meaningfully contribute to those frameworks. This asymmetry operates across three dimensions simultaneously. Infrastructure: AI compute, storage, and model development capacity is concentrated in five jurisdictions, with access for developing country practitioners mediated entirely through USD-denominated hyperscaler platforms. This is a governance outcome - produced by tax policy, R&D investment, and immigration frameworks in advanced economies - not a neutral market condition. Compliance: Regulatory frameworks for AI - impact assessments, transparency requirements, audit obligations - are calibrated for large-jurisdiction regulators and enterprise-scale deployers. Applied without structural adaptation to developing country SMEs, they impose compliance costs asymmetric with both the benefits received and the institutional capacity available. The WTO Investment Facilitation for Development Agreement (MC14, Yaoundé, March 2026) demonstrates that structured adaptation - not standards transfer - is what produces development-grade outcomes. Cultural economy: The creative and cultural sectors in developing countries are a distinct AI deployment context not captured in any listed theme. These sectors face specific pressures: training data skewed toward dominant cultural contexts, artist rights frameworks that predate AI, and capacity-building programs calibrated for tech enterprises rather than cultural producers. The Dialogue should establish a dedicated working stream on AI governance structural asymmetry, with specific attention to cultural economies in developing countries.

In Latin America's cultural economy, the governance gaps in our four priority areas produce a specific and measurable pattern of exclusion. The most significant challenge is infrastructure dependency. Avant.Dev deploys AI in cultural production workflows in Mexico City — generative tools for visual art, language models for policy intelligence, classification systems for digital archives. Every one of these applications runs on USD-denominated compute infrastructure located outside our jurisdiction. We have no access to locally trained foundation models built on Latin American cultural datasets. The governance gap here is not a technology gap — it is a capital allocation gap produced by the absence of any multilateral framework requiring AI infrastructure investment to be geographically distributed. The second challenge is compliance asymmetry. As AI regulatory frameworks mature in the EU and emerge across OECD jurisdictions, they cascade into developing country contexts as de facto requirements — without the institutional support structures that make compliance viable for large organizations in advanced economies. For a 22-artist collective in Mexico City, conformity assessment obligations designed for enterprise-scale deployers represent a structural barrier to legal AI adoption, not a protection. The third challenge is capacity-building calibration. Programs delivering AI skills training to Latin America consistently transfer frameworks designed for tech industry contexts. Cultural producers — artists, archivists, community media organizations — have fundamentally different data governance needs, different IP contexts, and different infrastructure constraints. Generic capacity-building does not reach them. The opportunities are real and directly tied to these gaps. Open-source models trained on locally curated cultural data would unlock AI applications for Latin American creative sectors that no hyperscaler is building. Interoperable governance frameworks with SME-differentiated compliance pathways would allow cultural economy practitioners to adopt AI legally and sustainably. These are not aspirational — they are actionable if the Dialogue designs for them.

The AI Dialogue occupies a unique position in the international governance architecture: it is the only process with a universal UN mandate, multistakeholder scope, and explicit thematic breadth across cultural, social, economic, and technical dimensions simultaneously. No existing bilateral, regional, or sectoral mechanism covers that full range. Its most valuable function in advancing international cooperation is not producing another set of principles. Enough principles exist. Its value lies in three specific cooperative roles that no other process is currently fulfilling. First, convergence mapping. The Dialogue can systematically document where governance frameworks across jurisdictions — the EU AI Act, the OECD AI Principles, the African Union AI framework, national strategies — converge and diverge, and what those divergences cost developing country practitioners in compliance terms. This is technical, actionable, and currently unmapped. Second, structural adaptation standards. International cooperation on AI governance has produced interoperability commitments primarily for advanced economy regulators. The Dialogue can establish the precedent that interoperability must include compliance pathway differentiation for developing country SMEs — a structural adaptation standard that multilateral bodies, including the IMF and World Bank, can incorporate into their own conditionality and technical assistance frameworks. Third, evidence integration from non-state practitioners. The Dialogue's multistakeholder mandate is its strongest asset and its least developed capability. Governments and large institutions produce the evidence base; SME cultural producers in developing countries are the affected constituency with the least representation in that base. The Dialogue can establish a practitioner evidence integration mechanism — a structured pathway for organizations like Avant.Dev to contribute operational data, not only written submissions — that feeds directly into the Co-Chairs' synthesis for the General Assembly. Cooperation without evidence from those most affected is not cooperation. It is consultation theater.

Several existing mechanisms provide the institutional infrastructure the Dialogue should integrate rather than replicate. WSIS Action Line C8 (Cultural Diversity, UNESCO): The most directly relevant existing framework for cultural economy AI governance. UNESCO has been facilitating C8 since 2003. The Dialogue should formally connect its Cluster 1 work to C8 implementation, tasking the UNESCO-led process with developing cultural economy AI impact metrics under the renewed WSIS mandate (A/RES/80/173, to 2035). This creates accountability continuity beyond Geneva 2026. ITU AI Readiness Framework 2.0 (January 2026): Dimension 4 — Contextualization and Regional Impact — is the most operationally developed methodology for assessing AI governance fitness across different economic contexts. The AI-RE Toolkit pilot at the July 2026 AI for Good Summit is a live implementation vehicle. The Dialogue should mandate that cultural economy parameters be embedded in the Toolkit methodology from the pilot stage. WTO Investment Facilitation for Development Agreement (MC14, Yaoundé, March 2026): The IFD Agreement's structured adaptation model — including EIB financing of up to EUR 1 billion and a WTO-China-ITC technical assistance framework — provides a replicable architecture for AI governance adaptation in developing economies. The Dialogue should reference IFD as a design precedent. Inter-Agency Working Group on AI (Para. 86, A/RES/80/173): Already mandated to report to this Dialogue. Its added value depends entirely on whether it incorporates practitioner evidence from developing country cultural economy actors — currently not embedded in its methodology. The added value the Dialogue uniquely brings: political authority to require that these mechanisms work together. Each initiative above operates in institutional silos. The Dialogue is the only process with the mandate and convening power to require coordination across UNESCO, ITU, WTO, and the UN Inter-Agency system simultaneously.

The current format — written submissions plus time-limited oral interventions — privileges organizations with policy writing capacity and English-language fluency. These are not proxies for relevance to AI governance outcomes. The format should be restructured across three dimensions. First, differentiated contribution pathways. Governments contribute policy positions. Academia contributes research. Practitioners — cultural producers, SME operators, community organizations — contribute operational evidence: what AI governance frameworks actually cost, where they fail, and what adaptation would require. The Dialogue should create a structured practitioner evidence track distinct from general civil society submissions, with dedicated synthesis in the Co-Chairs' report to the General Assembly. Second, regional preparation sessions with synthesis integration. Virtual consultations held in a single timezone-inclusive window do not generate the same quality of engagement as regionally prepared contributions. The Dialogue should mandate regional preparatory consultations — coordinated through existing bodies such as ECLAC for Latin America, UNECA for Africa, and ESCAP for Asia-Pacific — whose synthesis documents feed directly into the Geneva agenda, not as background documents but as primary inputs. Third, a structured response mechanism. The current format allows stakeholders to speak but not to respond to what they hear. A structured response round — even 60 seconds per stakeholder after the governmental segment — would transform the Dialogue from a parallel monologue format into an actual exchange. The APC coalition requested this in April 2026 consultations. It is technically simple and structurally significant. The Dialogue's added value over written submissions is its convening power. That power is wasted if the format produces the same output as an email inbox. Structure should be designed to generate synthesis, not just input.

Three constituencies are systematically underrepresented and structurally unlikely to self-correct without deliberate design intervention. Cultural and creative sector practitioners in developing countries. Artists, archivists, community media organizations, and cultural intermediaries are among the populations most directly affected by AI in their daily work — training data extraction without consent, algorithmic curation displacing local content, generative tools trained on cultural heritage without attribution or benefit-sharing. None of these communities have the institutional infrastructure to engage UN governance processes. They are not NGOs. They do not produce policy briefs. Their evidence exists in practice, not in submissions. Inclusion requires: dedicated practitioner panels at Geneva, stipended participation pathways for developing country cultural organizations, and synthesis mechanisms that convert operational testimony into policy-legible evidence. Indigenous communities and language minority groups. AI training datasets systematically underrepresent indigenous languages and knowledge systems — in some cases actively extracting them without consent. These communities face the most acute linguistic and cultural AI governance failures and have the least access to the processes that would address them. Inclusion requires: meaningful consultation with indigenous representative bodies before Geneva, not during it, and formal recognition of indigenous data sovereignty as a governance principle within Cluster 1. SME operators in developing economy digital sectors. The entities most affected by compliance cost asymmetries are small businesses — not civil society organizations with governance expertise, but operators trying to use AI tools legally in contexts where regulatory frameworks were designed for markets ten times their size. Inclusion requires: a dedicated SME evidence mechanism within the Inter-Agency Working Group on AI, and a compliance cost reporting protocol that captures developing country SME experience as primary data. Underrepresentation is not an access problem. It is a format problem. The format must change.

Three format innovations would materially improve the quality of engagement without requiring structural changes to the Dialogue's mandate or timeline. Practitioner testimony panels with synthesis integration. Modeled on the format used in parliamentary committee hearings and some IPCC working group sessions: structured testimony from practitioners — not prepared statements, but facilitated responses to specific questions from a panel moderator — with a designated synthesis rapporteur who converts the testimony into policy-legible findings within 48 hours. This format generates evidence. The current intervention format generates positions. Live evidence stress-testing. After the Scientific Panel presents findings, a dedicated session where practitioners from developing countries are invited to challenge, contextualize, or extend the data with operational evidence from their own contexts. Not a Q&A — a structured evidence dialogue, with the Scientific Panel required to respond substantively and log where its models do not account for developing economy conditions. This would be the first AI governance process to treat practitioner knowledge as data, not commentary. Structured carry-forward mechanism between consultations. The April 2026 virtual consultations and the July 2026 Geneva Dialogue are currently disconnected — there is no public record of how virtual consultation inputs shaped the Geneva agenda. A structured carry-forward document — published before Geneva, mapping each thematic area to the evidence submitted in consultations, with Co-Chair annotations on how that evidence was weighted — would transform stakeholder engagement from performative to accountable. Organizations would know whether their input reached the room. The goal is not more formats. It is formats that generate synthesis. The Dialogue should measure its own success not by the number of stakeholders consulted, but by how much practitioner evidence from developing countries appears in the final report to the General Assembly.

Three existing approaches offer concrete, replicable models for effective AI governance that the Dialogue should study and scale. The WTO Investment Facilitation for Development Agreement (MC14, Yaoundé, March 2026) demonstrates that structured adaptation - not standards transfer - produces measurable development outcomes. Its architecture: mandatory provisions calibrated for developing country institutional capacity, differentiated implementation timelines, and dedicated financing (EIB EUR 300 million initial tranche, up to EUR 1 billion mobilized). WTO research projects a 9 percent increase in global FDI and approximately 1 percent increase in global GDP from this structured approach. The design principle - adaptation built in from the start, not retrofitted - is directly applicable to AI governance frameworks. The ITU AI Readiness Framework 2.0 (January 2026) provides the most operationally credible methodology for context-sensitive AI governance assessment currently available. Its five dimensions - including Dimension 4 (Contextualization and Regional Impact) - allow governance readiness to be evaluated against local infrastructure, local data ecosystems, and local innovation capacity, not against a universal benchmark calibrated for advanced economies. The AI-RE Toolkit pilot at the July 2026 AI for Good Summit is a live implementation vehicle the Dialogue should monitor and formally connect to its own evaluation framework. At the practitioner level, Avant.Dev's own operational model offers a small-scale but replicable example of what culturally adapted AI governance looks like in practice. We deploy AI in cultural production workflows - generative tools for visual art, language models for policy intelligence - under a governance approach that prioritizes local data curation, artist rights documentation, and compliance transparency. We do this without enterprise-scale legal teams, which means our compliance architecture has to be structurally efficient. That constraint produces innovation. The Dialogue should create mechanisms to surface and learn from this kind of practitioner-generated governance knowledge systematically. Good practice is already happening at the margins. The Dialogue's role is to bring it to the center.