Africa Governance and Civic Innovation Hub

In automated decision making, discretion is not exercised at the moment of decision. It is embedded upstream at the point of system design, training, and configuration. By the time a government decision is produced through an AI system, the official nominally responsible has already been displaced from the moment of choice. Existing accountability frameworks are built for retrospective review of discrete decisions. They are structurally unsuited to governance of a process where the locus of power is prospective and distributed across technical actors operating outside the legal order. The Dialogue succeeds if it begins to develop doctrine for this reality not only transparency obligations and audit rights, but the legal infrastructure for what Africa Governance and Civic Innovation Hub have identified as Administrative Hosting Capacity: the institutional ability of a public authority to legally anchor, continuously supervise, recalibrate, and remain accountable for an AI system across its operational life, not only at the point of deployment. The second measure of success is structural inclusion. Africa is producing original governance doctrine not adapting frameworks designed elsewhere. The Dialogue succeeds if African institutions participate in the architecture of international AI governance, not merely in its implementation. A Dialogue that consults the Global South and codifies the Global North will have reproduced the governance inequality it claims to address.

• Interoperability of governance approaches
• Transparency, accountability, and human oversight
• AI capacity-building
• Safe, secure and trustworthy AI

My selection reflects priorities I have engaged not as a matter of policy preference but through original doctrinal research and active institutional work at Africa Governance and Civic Innovation Hub (AGCIH). Interoperability of governance approaches is the priority that frames all others. Through AGCIH's Governance Readiness Assessment Framework, I have developed a tool that evaluates whether a public authority has the institutional capacity to govern AI systems in continuous operation and not merely whether it has adopted a policy. Interoperability must be built at the level of governance structure, not regulatory text. States with different constitutional orders and administrative traditions must be able to converge on shared principles without inheriting a single model. Transparency, accountability, and human oversight are not procedural add-ons. They are the substance of governance. My research on Administrative Hosting Capacity ( original doctrine by AGCIH) establishes that oversight must be substantive and continuous not a moment of adoption but a sustained legal relationship between a public authority and the system it operates. Transparency without hosting capacity produces disclosure without accountability. AI capacity building, understood correctly, means institutional capacity, regulatory doctrine, legal anchoring frameworks, and the administrative competence to supervise automated systems. My current engagement with Zimbabwe's Judicial Service Commission on accessible justice governance makes this concrete: capacity is not only technical literacy. Safe, secure and trustworthy AI underpins the others. Trustworthiness is a governance outcome, not a product characteristic.

Two issues are absent from the listed thematic areas, and their absence reflects a structural gap in how AI governance is currently framed internationally. The first is procurement as the primary governance entry point. My research through Africa Governance and Civic Innovation Hub ( AGCIH) has identified what I term the Procurement Entry Doctrine: the moment at which a public authority acquires an AI system is the moment at which governance either begins or is forfeited. At procurement, the state delegates a governance function to a private actor operating under a commercial contract ,often before any regulatory framework applies. The legal consequences of that delegation, including questions of liability, recalibration rights, and exit, are almost entirely unaddressed in international governance discourse. The Dialogue should treat AI procurement governance as a distinct and urgent thematic area. The second is the accountability gap for AI deployed by international actors in state-adjacent spaces. During my work in Liberia on business and human rights, I observed directly how international organisations and development finance institutions operate governance-adjacent programmes with limited transparency and no enforceable accountability architecture. As these actors adopt AI in programme delivery, eligibility determination, and resource allocation, the accountability deficit deepens. The Dialogue must address not only how states govern AI, but how international actors deploying AI within states are held to equivalent standards. Both issues are cross-cutting. Neither can be resolved within any single thematic area listed in Resolution 79/325.

Zimbabwe and the broader African region are confronting a governance gap that is structural, not incidental. Public authorities across the continent are acquiring and deploying AI systems in justice administration, public service delivery, and regulatory functions without the institutional infrastructure to legally anchor, supervise, or recalibrate those systems over time. The gap is not primarily technical. It is doctrinal. Through the Africa Governance and Civic Innovation Hub (AGCIH) research and engagement with Zimbabwe's Judicial Service Commission and the Zimbabwe National AI Strategy, I have observed a consistent pattern: states are entering AI deployment through procurement decisions made before any governance framework exists. The Procurement Entry Doctrine I have developed through AGCIH identifies this as the foundational failure. Governance is forfeited at acquisition, before deployment begins. The consequences are concrete. Citizens subject to AI-assisted decisions in courts or administrative processes have no legal basis on which to challenge the upstream design choices that produced those decisions. Officials nominally accountable for those decisions have already been displaced from the moment of choice. The Relocation of Judgment ( developed by AGCIH) is not a theoretical concern in Africa. It is the daily operational reality of public administration adopting AI without hosting capacity; a doctrine named by AGCIH. The opportunity is equally real. Africa is at an early enough stage of AI deployment that governance architecture can be built prospectively rather than retrofitted. Several African governments are developing national AI strategies now and Zimbabwe's was launched on 13 March 2026. The doctrinal choices made in this window will shape a generation of governance practice. The Global Dialogue represents a genuine opportunity to ensure that African governance institutions participate in designing the international frameworks they will be expected to implement.

The Dialogue's most significant contribution to international cooperation would be to shift the register of cooperation itself. Current international engagement on AI governance operates primarily at the level of shared declarations and principle catalogues. What is absent is cooperation at the level of governance architecture, shared structural standards for how states build and maintain the institutional capacity to govern AI systems in continuous operation. International cooperation means very little if states cooperating on AI policy lack the institutional infrastructure to implement any commitment they make. Through AGCIH's engagements across Zimbabwe's justice and digital governance systems, I have observed this gap directly: governments adopting AI governance language without the administrative hosting capacity to give that language legal effect. Cooperation that does not address institutional capacity reproduces the gap it is designed to close. The Dialogue can advance genuine cooperation in three specific ways. First, by developing shared criteria for evaluating governance readiness; what AGCIH has formalised as the Governance Readiness Assessment Framework so that cooperation is calibrated to the actual institutional position of each state, not assumed. Second, by establishing that interoperability of governance frameworks is a structural goal, not a textual one: states need convergence on accountability mechanisms, oversight standards, and procurement governance principles, not uniform regulatory language. Third, by creating a peer review architecture through which states can assess and support one another's institutional capacity over time, rather than producing a single framework and assuming implementation. Cooperation without institutional architecture is aspiration. The Dialogue's value is in bridging that gap.

Three categories of existing initiative offer the Dialogue its most productive foundations, drawn from direct institutional experience. At the multilateral level, UNESCO's Recommendation on the Ethics of AI represents the most substantive normative foundation available. Through AGCIH's engagement with UNESCO in Zimbabwe, I observed directly that the implementation gap between the Recommendation's principles and states' institutional capacity to act on them is significant. The Dialogue should invest in that implementation architecture, not produce a parallel normative instrument. At the regional level, the African Union's Continental AI Strategy and the emerging national AI strategies across member states, including Zimbabwe's, launched on 13 March 2026 — represent governance design processes that are active, politically committed, and institutionally grounded. The Dialogue's added value is in connecting these national processes to one another and to a shared doctrinal framework, rather than superseding them with a global instrument that does not account for the continent's governance realities. At the sectoral level, the World Bank's engagement with judicial and public sector governance including the Global Justice Programme's current work with Zimbabwe's Judicial Service Commission on accessible justice demonstrates that AI governance intersects directly with institutional reform programmes already underway. The Dialogue should map these intersections explicitly and engage with development finance institutions as actors whose AI deployment in state-adjacent spaces must itself be subject to governance standards. The Dialogue's added value is connective and architectural. It does not need to begin from nothing. It needs to build structure across what already exists.

Meaningful stakeholder contribution to the Dialogue requires a prior condition that is rarely stated explicitly: participants must have the institutional capacity to translate dialogue outcomes into governance action. Representation without implementation capacity produces inclusion on paper and exclusion in practice. The first is a normative track, bringing together member states, international organisations, and multilateral bodies to advance framework development. The second is an institutional capacity track, convening governance practitioners, administrative law specialists, regulatory institutions, and applied governance researchers to develop the implementation doctrine that framework commitments require. These tracks must be formally connected, not parallel silos: the normative outputs of the first track should be tested against the institutional realities documented in the second before they are finalised. On structure, the Dialogue should resist the temptation to produce a single comprehensive outcome document. The governance challenges it must address are not uniform. A modular architecture in which thematic outputs are developed through specialist working groups and then integrated would produce more durable results than a plenary-driven consensus text. African governance institutions must be represented not only in plenary sessions but within the working groups where architectural decisions are made.

The most consistently underrepresented participants in global AI governance discussions are not communities in the sociological sense. They are institutional actors: the administrative law practitioners, regulatory officers, judicial officials, and governance researchers in the Global South who are responsible for implementing AI governance in practice and who are producing original doctrine to address problems that internationally prominent frameworks have not yet reached. This distinction matters. Current global AI governance discourse privileges two categories of voice: large technology jurisdictions with established regulatory infrastructure, and civil society organisations with international networks and funding. What is systematically absent is the practitioner-governance voice from African and other Global South institutions. The people who sit inside public authorities navigating, in real time, what it means to govern AI systems within constitutional and administrative law frameworks that were not designed for continuous automated administration. Through AGCIH's engagement with Zimbabwe's Judicial Service Commission, I have worked directly with judicial officers confronting these questions without any international governance framework that addresses their specific institutional position. Their experience is not anecdotal. It is governance data of the highest relevance to the Dialogue, and it is entirely absent from current international discussions. Inclusion requires two changes. First, dedicated representation for governance practitioners and applied researchers from African institutions not as civil society voices, but as governance architecture contributors. Second, a formal mechanism for incorporating practitioner doctrine developed outside the major AI jurisdictions into the Dialogue's normative outputs. Consultation is not inclusion, architectural contribution is.

The format innovation the Dialogue most needs is one that does not yet exist in international AI governance processes: governance stress testing. Current international forum formats , panel discussions, working group negotiations, written consultations are designed to produce textual outputs. They are poorly suited to testing whether those outputs are implementable. A governance stress test format would present participating states and institutions with a structured set of real AI governance scenarios drawn from actual public administration contexts ,automated decision-making in courts, AI-assisted welfare eligibility determination, algorithmic procurement systems and require participants to demonstrate, against their own frameworks, how accountability would function in each case. This is not a simulation exercise. It is a diagnostic process, structured around the same evaluative criteria as AGCIH's Governance Readiness Assessment Framework, that would reveal where international governance commitments are substantive and where they are aspirational. The outputs of that diagnostic would be more valuable to the Dialogue's normative work than any number of position papers. A second format innovation is structured doctrine exchange between governance institutions across jurisdictions not policy dialogue, but the direct exchange of legal frameworks, administrative instruments, and regulatory methodologies between institutions actively governing AI in public administration. AGCIH's current work on Administrative Hosting Capacity and the Procurement Entry Doctrine represents exactly the kind of developed governance methodology that such exchanges would surface. African institutions should be central contributors to that exchange, not peripheral observers. The Dialogue will produce better governance if it tests governance, not only discusses it.

The most instructive examples of effective AI governance practice share a common characteristic: they treat governance as an institutional design problem, not a policy drafting exercise. Three practices merit attention, the first drawn from AGCIH's own institutional work. AGCIH's Procurement Entry Doctrine and its associated AI Procurement Governance Addendum offer a practical governance instrument developed for public authorities in the African context. Rather than producing a general AI policy framework, the Addendum operationalises governance at the point of acquisition, the moment at which a public authority first assumes legal responsibility for an AI system's outputs. It establishes recalibration rights, transparency obligations, exit conditions, and accountability allocation within the procurement instrument itself, before deployment begins. This approach has been developed through active engagement with Zimbabwe's public administration and justice sector, and its design is replicable across different constitutional and administrative law contexts without requiring uniform regulatory frameworks. At the multilateral level, UNESCO's Recommendation on the Ethics of AI remains the most substantive normative foundation in the field. Its value lies not in its principles, which are broadly shared, but in its governance ethics impact assessment methodology, which provides states with a structured process for evaluating AI systems before deployment. Its limitation is implementation: few states have the institutional infrastructure to apply it systematically. Connecting the Recommendation's methodology to institutional capacity-building programmes would significantly increase its practical effect. At the national level, Zimbabwe's National AI Strategy, launched in March 2026, represents an example of a developing state making a governance-first commitment at an early stage of AI deployment. The design window in which governance architecture is most effective and least costly to build.