Alliance for Equitable AI

A successful first Global Dialogue on AI Governance would be one that moves beyond high-level principles and produces actionable, equity-driven outcomes, particularly for the Global South. First, the Dialogue must explicitly recognize and address AI Divides, especially the emerging Compute Divide. Bridging digital inequality today requires more than connectivity; it requires access to sovereign compute, localized datasets, and infrastructure that enables countries to meaningfully participate in AI development—not just consume it. Second, success would include a shift toward an AI Global Public Goods framework, with concrete pathways for compute-sharing agreements and open-access localized data ecosystems. Without this, interoperability risks becoming a one-directional process where developing countries adopt external standards without the capacity to innovate within them. Third, the Dialogue should embed linguistic and cultural equity into its outcomes. This includes mandating human-centric, localized AI validation within the work of the Scientific Panel, ensuring that AI systems are contextually relevant and not culturally misaligned. AI governance must avoid reinforcing forms of digital exclusion or "digital colonialism." Finally, success depends on continuity beyond a single event. The establishment of a Global South AI Observatory would be a meaningful outcome—tracking equitable access benchmarks, monitoring distribution of AI infrastructure and R&D, and ensuring that inclusion is measured in practice, not just principle. Ultimately, the Dialogue will be successful if it ensures that the Global South is not only represented in discussions, but empowered to shape the infrastructure and future of AI governance.

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

These priorities reflect the need to address AI governance from both a systems and equity perspective, particularly for emerging economies. AI capacity-building is foundational, but must extend beyond skills and literacy to include infrastructure capacity, particularly access to compute and data. Without this, meaningful participation in AI ecosystems remains limited. The social, cultural, and linguistic implications of AI are critical, as current systems often reflect biases rooted in non-representative datasets. Ensuring linguistic equity and cultural relevance is essential to prevent systemic exclusion and to build trust in AI systems across diverse populations. Interoperability of governance approaches is important, but must be approached cautiously. Without addressing underlying inequalities, interoperability may reinforce dependency, where developing countries are required to align with standards they had little role in shaping. Finally, open-source, open data, and open AI models are key enablers of equitable access. These can support localized innovation, reduce entry barriers, and allow countries to adapt AI systems to their own contexts. However, these must be supported by governance frameworks that ensure fair access and responsible use. Together, these priorities emphasize the need for inclusive, context-aware, and infrastructure-backed AI governance, rather than purely normative or principle-based approaches.

A key cross-cutting issue not explicitly captured is the Compute Divide as a structural driver of inequality in AI. While themes such as capacity-building and access are addressed, there is limited focus on who controls the computational infrastructure that underpins AI systems. Access to high-performance computing, cloud infrastructure, and large-scale data environments remains highly concentrated, creating systemic barriers for developing countries. Related to this is the issue of data sovereignty and localized data ecosystems. Many countries lack the infrastructure and governance mechanisms to generate, store, and utilize their own data effectively, which limits their ability to develop contextually relevant AI systems. Another emerging issue is the risk of digital colonialism, where AI systems developed in one context are deployed globally without sufficient adaptation to local social, cultural, and linguistic realities. This can reinforce existing power imbalances and reduce the agency of local stakeholders. Finally, there is a need for mechanisms that ensure continuity and accountability beyond dialogue processes. Without structured follow-up systems, such as observatories or monitoring frameworks, there is a risk that commitments remain aspirational rather than operational. Addressing these cross-cutting issues is essential to ensure that AI governance frameworks are not only inclusive in intent, but also equitable in outcome

Governance gaps in the selected thematic areas are already shaping both the risks and opportunities for countries like Pakistan and much of the Global South. The most significant challenge is structural inequality in access to AI infrastructure, particularly compute and high-quality localized datasets. While global discussions emphasize interoperability and governance frameworks, many countries lack the foundational capacity to participate meaningfully. This creates a growing Compute Divide, where nations are positioned primarily as consumers of externally developed systems rather than contributors to AI innovation. This imbalance is further reflected in the social, cultural, and linguistic implications of AI. Current models are often trained on non-representative data, resulting in systems that are misaligned with local contexts. This affects everything from language understanding to policy applications, raising concerns around trust, usability, and long-term digital sovereignty. Another governance gap lies in capacity-building, which is often framed around skills development but does not sufficiently address infrastructure readiness or institutional capabilities. Without strengthening both, governance frameworks risk remaining theoretical rather than implementable. At the same time, there are clear opportunities. The growing focus on open-source models, open data, and global collaboration presents a pathway for more inclusive participation, if supported by equitable access mechanisms. There is also increasing recognition of the need to address AI divides at a structural level, which opens space for new models such as compute-sharing frameworks and localized AI ecosystems. For countries like Pakistan, the challenge—and opportunity—is to move from passive adoption to active participation, ensuring that AI governance frameworks enable not just compliance, but meaningful inclusion in shaping the future of AI systems.

The AI Dialogue can play a critical role as a neutral, multilateral platform that bridges fragmented global efforts on AI governance while ensuring that cooperation is both inclusive and actionable. First, it can help move the conversation from principles to implementation, by translating high-level commitments into practical frameworks that countries can adopt based on their capacities. This is particularly important for the Global South, where governance discussions often outpace infrastructure readiness. Second, the Dialogue can serve as a platform to rebalance participation, ensuring that developing countries, civil society, and underrepresented stakeholders are not only present but are able to meaningfully shape outcomes. This includes addressing structural barriers such as limited access to compute, data, and technical resources, which directly impact the ability to engage in global AI ecosystems. Third, the Dialogue can advance cooperation by promoting shared resources and collective models, such as global public goods approaches to AI. This could include facilitating collaboration on open datasets, compute-sharing mechanisms, and interoperable governance tools that reduce duplication and enable broader participation. Finally, the AI Dialogue can play a convening role in establishing continuity and accountability mechanisms, ensuring that discussions lead to sustained collaboration rather than one-off engagements. By creating structured pathways for follow-up, it can help align international efforts and support a more coordinated global response to AI governance challenges. In this way, the Dialogue has the potential to act not just as a forum for discussion, but as a catalyst for equitable and cooperative AI governance globally.

The AI Dialogue should build upon and connect with existing global and multilateral initiatives, including efforts led by UN agencies (such as ITU, UNESCO, and UNICEF), the Global Digital Compact, the AI for Good Global Summit, and regional and national AI strategies. It should also draw from ongoing work by academic institutions, civil society, and industry-led partnerships that are already contributing to AI governance discussions. While these initiatives provide valuable foundations, they often operate in parallel or fragmented ecosystems, with varying levels of participation and accessibility for developing countries. The added value of the AI Dialogue lies in its ability to create coherence across these efforts, bringing them into a more unified and inclusive framework. A key contribution of the Dialogue would be to ensure that these existing initiatives are not only connected, but also contextualized for diverse national realities, particularly in emerging markets. This includes integrating perspectives from countries that face structural barriers such as limited infrastructure, data ecosystems, and institutional capacity. Additionally, the Dialogue can add value by introducing equity-focused mechanisms, such as frameworks for open access, localized data development, and shared infrastructure models. By aligning existing initiatives with these priorities, it can help shift the focus from isolated innovation to collective, inclusive progress. Ultimately, the AI Dialogue's role is to act as a convergence point—one that not only connects existing efforts, but also strengthens them by embedding principles of equity, accessibility, and practical implementation across the global AI governance landscape.

Different stakeholders can contribute meaningfully to the AI Dialogue if the format moves beyond sequential interventions toward structured, interactive engagement. Governments can provide policy direction and regulatory insights; academia can contribute research and evidence; industry can share practical implementation challenges; and civil society can ground discussions in lived realities and equity concerns. However, to unlock this diversity, the Dialogue must ensure that contributions are not limited to short, isolated statements. A more effective structure would include: Thematic working sessions with guided questions, allowing stakeholders to respond to each other in real time Balanced speaking allocation, ensuring meaningful participation from non-governmental actors, especially from the Global South Pre-consultation inputs, where stakeholders submit written contributions that are synthesized and used to inform discussions Outcome-oriented sessions, focused on identifying actionable recommendations rather than broad statements Additionally, the Dialogue should incorporate mechanisms for continuity, such as working groups or follow-up platforms, to ensure that contributions are carried forward beyond the event. Ultimately, stakeholder engagement will be most effective if the Dialogue is designed not just as a forum for expression, but as a space for collaborative problem-solving and co-creation of solutions.

Despite growing participation, several critical voices remain underrepresented in global AI governance discussions. These include: Stakeholders from the Global South, particularly those from low and middle-income countries with limited access to AI infrastructure Local developers, startups, and technical communities working on context-specific solutions Non-English speaking communities, whose linguistic realities are often not reflected in current AI systems Grassroots organizations and civil society groups, especially those working with vulnerable populations such as refugees, women, and rural communities Public sector implementers (e.g., judiciary, health systems), who face real-world challenges in deploying AI without adequate governance frameworks To include these voices, the Dialogue must address both access and structural barriers. This includes: Providing financial and technical support for participation Enabling multilingual engagement and localized consultations Creating regional and thematic entry points, where stakeholders can contribute in more accessible formats Recognizing and integrating on-ground experience and non-traditional expertise into decision-making processes Inclusion should not be limited to representation, but should ensure that these perspectives actively shape outcomes, particularly in areas related to infrastructure, access, and equity.

To foster meaningful and dynamic engagement, the AI Dialogue should move beyond traditional panel formats and adopt more interactive and participatory models. Some effective formats could include: Scenario-based exercises: Stakeholders engage with real-world AI governance challenges (e.g., health systems, judiciary, financial systems) to explore trade-offs and decision-making in practice Governance sandboxes: Small, diverse groups collaborate to test policy ideas, frameworks, or use cases in a controlled environment Thematic roundtables with mixed stakeholders: Bringing together governments, industry, academia, and civil society in structured discussions to co-develop recommendations Live polling and feedback mechanisms: Allowing participants to respond in real time, helping identify areas of consensus and divergence Pre-recorded interventions + live discussion: Reducing time constraints while enabling deeper engagement during sessions Additionally, the Dialogue could introduce "continuity tracks", where participants engage over time through digital platforms, working groups, or collaborative spaces. This would allow ideas to evolve beyond a single event. Importantly, formats should be designed to encourage dialogue, not just delivery—ensuring that participants can build on each other's insights rather than present in isolation. By adopting such approaches, the AI Dialogue can become a space for active collaboration and practical problem-solving, rather than a series of disconnected interventions.

Effective AI governance requires a combination of policy frameworks, technical standards, and collaborative platforms that are both inclusive and implementable. At the policy level, initiatives such as UNESCO's Recommendation on the Ethics of Artificial Intelligence and emerging national AI strategies provide important normative guidance. However, their impact is strengthened when paired with practical implementation mechanisms, such as regulatory sandboxes and sector-specific governance models that allow controlled experimentation and adaptation. From a technical and operational perspective, open-source AI ecosystems (e.g., open models, shared datasets, and collaborative development platforms) offer a pathway toward more equitable participation. These approaches lower entry barriers and enable countries and local developers to build context-specific solutions, particularly when supported by responsible governance frameworks. There is also growing value in multi-stakeholder platforms such as the AI for Good Global Summit and similar initiatives, which bring together governments, industry, and civil society to exchange knowledge and align efforts. These platforms are most effective when they move beyond dialogue to facilitate resource-sharing, capacity-building, and joint problem-solving. A key emerging approach is the concept of AI as a Global Public Good, which emphasizes shared access to critical resources such as compute infrastructure and localized data ecosystems. Mechanisms such as compute-sharing frameworks and open-access data initiatives can help address structural inequalities and enable broader participation in AI development. Finally, localized validation and human-centric design practices are essential to ensure that AI systems are culturally and contextually relevant. Embedding these into governance processes can help build trust and avoid unintended exclusion. Together, these examples highlight that effective AI governance is not only about setting principles, but about creating accessible systems, shared infrastructure, and collaborative mechanisms that translate those principles into practice.