Rhodes University

A successful first Global Dialogue on AI Governance should move beyond symbolic consensus and produce clear, actionable, and context-responsive outcomes that reflect the diversity of global realities. First, it should establish a shared set of guiding principles for AI governance that foreground equity, accountability, transparency, and human dignity. These principles must explicitly recognise the risks of reproducing existing global inequalities, particularly for underrepresented languages, knowledge systems, and communities in the Global South. Second, the Dialogue should result in concrete commitments to inclusive participation in AI development. This includes mechanisms to support low-resource contexts through equitable data representation, investment in local language technologies, and capacity-building initiatives that enable meaningful participation in AI design, governance, and deployment. Third, a successful outcome would include the creation of a collaborative global framework or working groups tasked with addressing key governance challenges such as data sovereignty, algorithmic bias, and ethical AI use in education and public sectors. These structures should ensure ongoing dialogue rather than a once-off engagement. Fourth, the Dialogue should prioritise interdisciplinary and cross-sector collaboration, bringing together policymakers, researchers, educators, technologists, and community stakeholders. Governance must be informed not only by technical expertise but also by social, cultural, and linguistic perspectives. Finally, success would be measured by the establishment of accountability and follow-through mechanisms, including timelines, monitoring structures, and platforms for continued engagement. Without implementation pathways, even the most well-articulated principles risk remaining aspirational. Ultimately, the Dialogue would be successful if it shifts AI governance from being predominantly shaped by a few dominant actors to a more inclusive, plural, and contextually grounded global effort that ensures AI serves all, rather than reinforcing existing divides.

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

My selection reflects a priority for ensuring that artificial intelligence development and governance are inclusive, equitable, and contextually responsive, particularly in multilingual and under-resourced settings. The first selection is central, as AI systems do not operate in neutral contexts. They shape and are shaped by existing power relations, often privileging dominant languages and knowledge systems while marginalising others. Addressing these implications is critical to ensuring that AI does not reproduce or deepen existing inequalities. AI capacity-building is equally important, as meaningful participation in AI ecosystems requires more than access to technology. It involves developing the skills, knowledge, and institutional support necessary for educators, researchers, and communities to engage critically with AI, adapt it to local contexts, and contribute to its development. Without this, many regions risk remaining consumers rather than co-creators of AI technologies. The protection and promotion of human rights provides a necessary normative foundation for AI governance. This includes safeguarding linguistic rights, cultural representation, and equitable access to knowledge. In contexts where certain languages and communities are already underrepresented, AI systems must be intentionally designed to uphold these rights. Finally, transparency, accountability, and human oversight are essential to ensure that AI systems can be scrutinised and governed responsibly. This is particularly important in addressing algorithmic bias, improving trust, and enabling stakeholders to understand how decisions are made.

One critical cross-cutting issue is the digital and epistemic marginalisation of low-resource languages. While several themes touch on ethics, inclusion, and human rights, there is insufficient explicit focus on how AI systems systematically exclude languages that lack large digital datasets. This exclusion is not only technical but also epistemic, as it limits whose knowledge is represented, accessible, and legitimised in AI-mediated environments. Without deliberate intervention, AI risks reinforcing a hierarchy of languages where dominant global languages continue to shape knowledge production, while indigenous and local languages remain digitally invisible. Addressing this requires targeted investment in language data development, inclusive model design, and policies that prioritise linguistic diversity as a core component of AI governance, rather than a peripheral concern.

In South Africa, governance gaps in AI are already visible across everyday digital use, education systems, and broader institutional frameworks. At the level of social platforms, learners increasingly interact with AI-driven tools embedded in platforms such as Meta (e.g. automated translation, content recommendation, and generative assistants). While these tools expand access, they predominantly operate in English and other high-resource languages, often misrepresenting or excluding indigenous languages such as isiXhosa. This shapes how learners consume and produce knowledge, reinforcing linguistic hierarchies from an early stage. Within higher education, the rapid uptake of generative AI tools has outpaced institutional policy development. Universities are still grappling with questions of academic integrity, ethical use, and pedagogical integration. There is limited guidance on how AI can be meaningfully incorporated into multilingual teaching and learning, particularly in programmes preparing teachers for linguistically diverse classrooms. This creates uncertainty for both educators and students, while also risking uncritical adoption of tools that are not contextually responsive. At a broader governance level, South Africa faces challenges related to regulatory clarity, data governance, and equitable access. There is a lack of coordinated national frameworks that address AI in relation to language diversity, education, and social equity. Additionally, infrastructural inequalities, such as uneven access to devices, connectivity, and digital resources, further compound these challenges. However, these gaps also present significant opportunities. South Africa is well positioned to lead in contextually grounded AI development, particularly through investment in African language technologies, inclusive data ecosystems, and capacity-building initiatives. Strengthening collaboration between government, academia, and industry could enable the development of AI systems that are not only innovative, but also linguistically inclusive and socially responsive. Ultimately, addressing these governance gaps is critical to ensuring that AI contributes to educational transformation rather than reinforcing existing inequalities.

The AI Dialogue can play a crucial role by creating a shared, inclusive platform for coordination and co-creation of AI governance approaches across diverse contexts. It can bridge divides between the Global North and South by ensuring that governance frameworks are informed by varied social, linguistic, and economic realities, rather than being universally imposed. A key role of the Dialogue is to facilitate knowledge exchange and capacity-building, enabling countries with limited resources to meaningfully participate in AI development and regulation. It can also support the development of common principles and adaptable guidelines that promote ethical, transparent, and accountable AI while allowing for contextual flexibility. Importantly, the Dialogue can strengthen multi-stakeholder collaboration, bringing together governments, academia, industry, and communities to address shared challenges such as bias, data governance, and inclusion. Ultimately, it can shift AI governance towards a more equitable, cooperative, and globally representative model.

The AI Dialogue should build on existing initiatives that are already engaging with AI, language, and education in contextually grounded ways, particularly within African and multilingual settings. These include emerging African language natural language processing (NLP) initiatives, university-based research programmes in teacher education and digital literacy, and platforms that promote multilingual education and language intellectualisation in South Africa. In addition, collaborations between academia, civil society, and technology labs working on inclusive data and language technologies provide a strong foundation.

Different stakeholders should contribute through structured, multi-level participation. Governments can provide policy direction, academia can contribute research and evidence, industry can share technical expertise and infrastructure, and communities can foreground lived realities and contextual needs. The AI Dialogue should be organised into thematic working groups, supported by regional consultations and sector-specific forums (e.g. education, language, public services). It should include ongoing cycles of engagement, not a once-off event, with clear feedback loops.

Global discussions on AI governance continue to underrepresent speakers of low-resource languages, educators and learners in under-resourced contexts, grassroots communities, and researchers from the Global South. In particular, perspectives from African language communities are often absent, despite being directly affected by AI systems that shape communication, learning, and knowledge access. Educators, especially those working in multilingual classrooms, are also overlooked, even though they are key mediators of how AI is introduced and used in practice. These gaps are not only about participation but about epistemic representation, whose knowledge, language, and lived experiences are recognised in shaping AI systems and policies. Inclusion requires more than symbolic representation. First, there must be intentional resourcing and capacity-building to enable meaningful participation, including funding for researchers, institutions, and community-based organisations in underrepresented regions. Second, processes should support multilingual engagement, allowing contributions in diverse languages rather than privileging English as the default medium. Third, the Dialogue should incorporate practice-based voices, such as teachers, students, and local innovators, through structured platforms like regional consultations and sector-specific forums. Additionally, partnerships with universities, community organisations, and local technology initiatives can help surface grounded perspectives and ensure continuity beyond formal events.

At a policy level, frameworks such as the European Union's AI Act demonstrate how governance can incorporate risk-based regulation, transparency, and accountability, particularly for high-impact sectors like education. Similarly, the UNESCO Recommendation on the Ethics of Artificial Intelligence (2021) provides guidance on human rights, inclusion, and cultural diversity, emphasising that AI systems must respect linguistic and cultural contexts. In practice, initiatives in African language natural language processing (NLP), such as the development of datasets and models for low-resource languages, offer concrete solutions to the challenge of linguistic exclusion. These efforts show the importance of collaborative, open, and locally grounded innovation. In the education sector, emerging practices include the co-development of AI tools with educators, ensuring that technologies align with pedagogical needs and local realities. This participatory approach enhances both relevance and accountability.