ACRC (Africa Cybersecurity Resource Center)

Three criteria, in my view, distinguish a Dialogue that will have mattered from one that will simply have taken place. First, that the Dialogue explicitly recognise public communication as a dimension of AI governance, rather than a post-decisional deliverable. Without a communication infrastructure connecting decisions to the populations they concern, governance reduces to technical administration. Second, that the voice of Global South States and populations francophone Africa in particular be integrated into the process before Geneva, not only in July. Third, that the Dialogue equip itself with a multilingual communication dispositif for its own conclusions, designed upstream, addressed to non-expert audiences in the Global South. These three points are developed in the written submission attached.

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

These four priorities form a coherent line: they address the conditions under which AI governance becomes effectively democratic for the populations it concerns. Transparency, accountability and human oversight is the cluster where the central argument of my submission lands. Transparency, to be operational, presupposes a public communication infrastructure that renders governance decisions intelligible to non-expert citizens. Without it, transparency exists on institutional websites and in expert circles, not in public deliberation. Social, economic, ethical, cultural, linguistic and technical implications of AI matters here because the linguistic dimension is decisive. AI public policies in francophone Africa are designed in French, communicated in French through institutional channels, and remain inaccessible to citizens whose primary language of daily life is Wolof, Ewe, Mooré, or any of dozens of other national languages. A governance that does not address linguistic conditions of intelligibility cannot claim to be inclusive. AI capacity-building is the cluster where the most operational proposal of my submission belongs: a principle of "communicational readiness" for national AI strategies. Strategies should not be considered complete at publication. They should be accompanied by a sustained public communication dispositif reference narrative, citizen feedback mechanism, multilingual outreach designed as an integral part of the strategy itself. Protection and promotion of human rights anchors the entire argument. The right to take part in public decisions affecting one's life presupposes the right to understand those decisions. AI raises this stake to an unprecedented level because of its opacity and its scale. Human rights protection is hollow without communicational sovereignty. These four priorities, taken together, name the same gap from four angles. My written submission develops each one through concrete operational proposals.

Yes one cross-cutting dimension, in my view, is missing from the seven themes listed: public communication as an operational dimension of AI governance. The listed themes transparency, human oversight, capacity-building, linguistic implications touch the issue at its edges, but none of them treats it head-on. The result is a framing in which communication is implicitly understood as a post-decisional deliverable: press releases, institutional websites, occasional awareness campaigns. This conception is no longer sufficient for AI. The reason is precise. AI is already modifying the material conditions under which meaning is produced and circulated in contemporary societies. A governance that does not equip itself with its own communication infrastructure leaves room for commercial platforms to define, in everyday interfaces, what AI means for the ordinary citizen. Communicational sovereignty becomes a dimension of sovereignty itself. This gap is observable empirically. When Global South States publish ambitious national AI strategies Senegal 2023, Togo under way, Benin, Côte d'Ivoire the documents are posted online, debated by experts on professional platforms, and translated into donor-funded implementation projects. The populations concerned remain outside the loop. The ecosystem closes upon a restricted circle: State, technical experts, international donors. Few mechanisms ensure sustained communication towards non-expert citizens, in the languages they actually speak. The risk is that this national pattern reproduces itself at the multilateral scale. A Dialogue that mobilises institutional stakeholders without thinking explicitly about its communication towards Global South populations will be perceived, in many regions, as one more exercise among insiders. Public communication, treated as governance infrastructure rather than as an accessory, deserves to be named explicitly either as a cross-cutting principle or as an operational condition under cluster 4. My written submission develops this argument and proposes four concrete operational steps.

The governance gaps in the four areas selected manifest concretely in West African francophone digital ecosystems, where the author works.Linguistic and communicational gap. Senegal's 2023 National AI Strategy, Togo's strategy under elaboration, and comparable documents in Benin and Côte d'Ivoire are designed and communicated in French, through institutional channels. The francophone state administration speaks French; the citizen of Saint-Louis, Sokodé, or Korhogo speaks Wolof, Ewe, Senoufo or any of dozens of national languages used in daily life. The linguistic chain that should connect strategy to citizen breaks at the first step.Capacity-building gap. Real implementation projects exist: the Goin' Digital programme in Senegal aims to train 90,000 people in data science by 2028; Togo's partnership with Kira Learning addresses programming, AI and English in public universities. These efforts are genuine. They are not accompanied, however, by a sustained public communication dispositif explaining what these initiatives concretely change for citizens outside the targeted training pools. Human rights and oversight gap. The Novissi project in Togo, widely cited internationally, used predictive algorithms on geospatial data to target 57,000 vulnerable citizens for cash transfers. The publicly available documentation does not evidence a prior explanatory communication directed at the targeted populations regarding their identification, the data used, or available recourse. A successful technical project does not automatically constitute a successful governance project.The opportunity is precise. Francophone Africa has the political will, the institutional momentum, and a growing pool of trained communication and AI professionals. What is missing is a recognised framework that treats public communication as governance infrastructure not as an accessory. Such a framework, if articulated by the Global Dialogue, would find immediate practical relevance in this region.My written submission develops these elements.

The Dialogue can play a role distinct from any of the existing forums on AI governance, on three conditions. A function of articulation, not duplication. The multilateral landscape on AI governance is already dense: GPAI, OECD AI Principles, UNESCO Recommendation on AI Ethics, Council of Europe Framework Convention, Global Digital Compact, Internet Governance Forum, regional initiatives. The Dialogue's specific value is not to add another normative layer. It is to articulate these initiatives surface convergences, name divergences, identify gaps and to do so in a universal UN-anchored space that no other forum offers. A function of inclusion that goes beyond declared intent. "Multistakeholder" and "inclusive" are widely used terms in this field. They are rarely operationalised. The Dialogue can demonstrate practical inclusion by ensuring that voices from Global South States, francophone Africa among them, are integrated into substantive preparation rather than into ceremonial participation. The credibility of the Dialogue's outputs in regions outside the traditional centres of AI governance depends on this. A function of communicational mediation. The technical and policy debates on AI are progressing rapidly within specialised circles. Public understanding of these debates lags far behind, including in OECD countries, and dramatically so in Global South contexts. The Dialogue is uniquely positioned, given its UN universality and its joint coordination by ITU, UNESCO and ODET, to produce a multilingual public communication output that translates institutional conclusions into intelligible content for non-expert audiences. This is not a public relations exercise. It is a governance function. A Dialogue that succeeds on these three fronts will have established something durable: a recognised universal space where AI governance is articulated, inclusively prepared, and publicly communicated. A Dialogue that succeeds on only one or two will have repeated patterns that already exist elsewhere. My written submission develops these points.

The Dialogue should build on initiatives organised around three complementary functions.Normative anchoring. The UNESCO Recommendation on the Ethics of AI (2021), the Council of Europe Framework Convention on AI and Human Rights (2024), and the OECD AI Principles already provide a substantial normative base. The Dialogue is not in a position to renegotiate these foundations and should not attempt to. It should reference them as starting points and articulate them with one another, including where their underlying philosophies differ.Operational and policy coordination. GPAI, the AI Safety Institutes network, the EU AI Act implementation work, and emerging national AI strategies including in Senegal, Togo, Rwanda, Kenya, Singapore generate practical lessons that no single forum currently consolidates. The Dialogue can become the space where these lessons are surfaced, compared and made available to States that are still designing their own frameworks.Multistakeholder and Global South connection. The Internet Governance Forum, the WSIS process, Smart Africa, the African Union Continental AI Strategy, and UNESCO regional offices already convene voices that the Dialogue must integrate. Connecting with these structures, rather than addressing them only at high-level segments, will determine the Dialogue's actual representativity. The added value the Dialogue can bring lies in three places: a UN universality no other forum offers; a coordination capacity through the joint ITU-UNESCO-ODET secretariat that bridges technical, ethical and capacity-building communities rarely connected elsewhere; and the possibility of producing a public communication output that translates institutional consensus into intelligible content for populations outside specialised circles. This last function public communication as governance infrastructure is currently performed by no existing initiative. It is the most distinctive contribution the Dialogue can make.My written submission develops these points.

On stakeholder contributionStates should arrive in Geneva with concrete national experiences to share, not only with positions to defend. The richest moments of any multilateral process are those where a Minister or Ambassador describes what worked and what failed in their own jurisdiction. The Dialogue should explicitly invite this register.Civil society and academia should be present beyond high-level interventions. The thematic discussions need genuine multistakeholder time at minimum 40% of speaking slots in each cluster with non-State voices able to respond to State positions, not only to be heard sequentially.Practitioners working on AI implementation in Global South contexts public sector, NGOs, technical communities should be specifically resourced to attend. Their absence at Geneva would reproduce, at the multilateral scale, the gap the Dialogue is meant to address. On format and structure Day 1 thematic clusters should each conclude with a concrete output formulated in plain language, not only with a co-chair summary in technical register. This output should answer one question: what changes for citizens after this discussion? Day 2 plenaries should preserve sufficient unstructured time for cross-cluster connections to emerge. The strongest insights from preparatory consultations have been those that cut across themes; the Geneva format should not foreclose this.A dedicated francophone consultation, organised before Geneva in partnership with OIF and UNESCO regional offices in Africa, would correct an under-representation observable in current preparations. This consultation could feed substantively into the cluster on capacity-building and the cluster on linguistic implications.The two-day structure is workable. Its success depends less on its architecture than on whether speaking time and substantive preparation are distributed in ways that actually surface non-institutional voices.My written submission develops these points.

Three categories of voices are systematically underrepresented in current global AI governance discussions. Non-anglophone administrations and civil societies of the Global South. Francophone Africa is one example, observable in the preparations of this Dialogue itself: dozens of countries, hundreds of millions of speakers, public administrations operating in French, and not a single dedicated consultation in the announced preparatory programme. The same observation applies to lusophone Africa, the Spanish-speaking Caribbean, and several non-English Asian contexts. Inclusion requires more than translation of final outputs. It requires substantive consultations conducted in working languages other than English, with regional partners organising them.Non-expert citizens directly affected by AI deployments. Algorithmic targeting in social protection programmes, automated decisions in administrative procedures, AI-mediated public services these affect populations who are rarely heard in governance discussions, even when they are the primary subjects of the decisions. Inclusion here means investing in deliberative formats citizen panels, structured public consultations, accessible explanatory materials in national languages rather than relying on civil society organisations to speak on their behalf.Practitioners outside the policy-research-industry triangle. Strategic communication professionals, public administration officers, educators, journalists working on AI in their daily practice carry operational knowledge that policy and academic discussions rarely capture. They are not absent because they are uninterested. They are absent because the formats are not designed for them. Inclusion requires explicit slots for practitioner voices in thematic discussions, not only auxiliary status at side events.A common thread runs through these three categories: representation depends less on declared openness than on the concrete design of access language, format, resources, timing. The Dialogue's credibility on inclusion will be measured against this design, not against the rhetoric surrounding it.My written submission develops these points.

Four formats deserve consideration, each addressing a different limitation of standard multilateral engagement. Structured response sessions, not sequential interventions. The standard format three-minute statements followed by no exchange produces a record but rarely produces dialogue. A more demanding format would allocate, in each thematic cluster, a session where two or three pre-identified positions are presented, then directly responded to by stakeholders from a different category (a State position responded to by civil society, an industry position responded to by Global South practitioners). The friction is the point. Practitioner reading panels. A short slot in each cluster where two or three practitioners public officials, NGO operators, professionals working on AI implementation read a single concrete case from their experience: what they tried, what worked, what did not. Two minutes per case, no abstractions. The rest of the cluster builds on this material rather than on principles. A Global South listening session. A dedicated session, conducted partly in non-English working languages with simultaneous interpretation, where Global South voices speak first and the rest of the room responds. Reversing the standard order surfaces priorities that often disappear when they have to wait for high-level segments to conclude. A public communication design workshop. A working session, civil society and communication professionals together, dedicated to drafting the Dialogue's own public communication output: how the conclusions of Geneva will be translated and circulated to non-expert audiences. Treating this as a substantive workshop, not an editorial afterthought, would model the principle the Dialogue is asked to recognise. These four formats share one principle. Engagement is not determined by who is present. It is determined by how speaking time is structured.My written submission develops these points.

Four examples illustrate distinct approaches that the Dialogue can build upon, each addressing a different challenge. Mauritius National AI Strategy 2025-2030 and FAIR Guidelines. Presented in April 2026, the Mauritian framework explicitly articulates ethical principles (fairness, accountability, inclusiveness, responsibility) with implementation mechanisms a regional AI marketplace, regulatory sandboxes, citizen-centric digital services. It demonstrates that small States can produce coherent national frameworks that are not reduced versions of OECD models. The document is publicly available and its implementation is being tracked. Singapore Model AI Governance Framework, second edition. Singapore offers an example of governance designed for operationalisation: principles translated into concrete checklists, sector-specific guidance, an independent AI Verify testing toolkit. The added value for the Dialogue is the framework's deliberate readability for practitioners public officials, sector regulators, mid-sized organisations rather than only for legal specialists.UNESCO Readiness Assessment Methodology (RAM). Conducted in over 60 countries, the RAM provides a structured methodology for States to assess their AI governance readiness across legal, institutional, technical, and societal dimensions. It is one of the few existing instruments designed specifically for capacity-building rather than for normative alignment. Open data platforms with public communication design. France's data.gouv.fr, the European Open Data Portal, and increasingly several African initiatives such as Sénégal Numérique's data platform, demonstrate that effective governance depends on public-facing infrastructure that is intelligible, navigable, and continuously updated. These platforms are not neutral repositories. They are governance infrastructure in the sense this submission defends. The common element in these four examples is that they treat governance as ongoing practice with public-facing components, rather than as one-off regulatory production. The Dialogue would do well to surface and connect such practical instruments. My written submission develops these points.