Laws.Africa

The success of the Global Dialogue will depend on its reach. AI development and implementation affect everyone, yet most of these conversations remain overtly technical and accessible only to academics or industry experts. A meaningful outcome requires that ordinary people (without technical expertise) understand what AI is, how it works, and how it will shape their futures. The Dialogue must unpack the original purpose and intention behind creating AI so that we can better identify and prevent misuse. For example, what is the societal value of a tool that allows image manipulation of a person without their consent? Could facial detection not be built in to prevent it? In terms of outcomes, a successful outcome must deliver an internationally accepted distinction between machine intellectual property and human intellectual property. Does AI-generated output belong to the AI system, its creators, or the human who prompted it? Where a human refines that output, or where AI demonstrably mimics an existing creative's work or persona, this must be factored into any framework. Equally critical is globally accepted accountability, i.e. clear standards and meaningful consequences. Both the user who generates harmful content and the creators and owners of the tools must be held responsible. Given the billions in revenue these systems generate, a higher burden of responsibility on developers and owners is entirely justified. Open-source AI models and open datasets must be a non-negotiable outcome, ensuring that developing nations and ordinary citizens are not permanently dependent on proprietary systems controlled by a handful of corporations. Transparency must also be built into any global framework, i.e. there must be an obligation to disclose when content is AI-generated, and creators of AI systems must declare what training data was used to build them. Finally, the environmental impact of AI's vast energy demands cannot be overlooked. There is no justification for the current pace of development. Human life is not threatened by slowing down, it is threatened by rushing ahead unchecked.

• Transparency, accountability, and human oversight
• Open-source software, open data and open AI models
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Safe, secure and trustworthy AI

I will explain my selection by sharing my reasoning for not choosing the remaining three priorities. I did not select "AI capacity-building" because I have observed that a wealth of knowledge and skills already exists across most nations. AI capacity therefore depends far more on the will of governments to ensure digital equality and localised infrastructure. In my opinion, the genuine gaps are addressable through the open-source priority. I did not select "Interoperability of governance approaches" though I would have liked to. Limited to four choices, I made a deliberate decision to prioritise foundational issues. I believe interoperability is a natural outcome of the conversations already taking place globally. Most, if not all, nations appear aligned on the fundamental requirements of AI governance, and a degree of interoperability will follow organically from that consensus. I did not select "Protection and promotion of human rights" because I believe that this priority has always been, and must remain, a responsibility across the board of those we elect and allow to occupy positions of leadership. Further, it is clear to all that AI amplifies the need for human rights protections, as it creates new categories of rights violations. The four priorities I have selected form what I consider the core of AI governance. Critically, I believe they are broad and interconnected enough to touch upon, and in many instances directly serve, the three priorities I did not select. Capacity-building is advanced through open-source access and transparency. Interoperability follows from agreed accountability standards. And human rights are best protected when AI systems are safe, trustworthy, and subject to genuine human oversight.

Yes. One critical cross-cutting issue that is not receiving the attention it deserves is the significant and lasting unemployment, and by extension, poverty, that AI advancement will inevitably cause. Most discussions trivialise this risk. The assertion that AI will not replace humans, but only those who fail to adapt and upskill, is not only simplistic, it is insensitive to the very real differences between people. Not everyone is technologically literate. Not everyone is capacitated to retrain and reposition themselves in an AI-driven economy. Digital inequality is already vast, and by the time meaningful efforts to close that gap bear fruit, large portions of the population may already be beyond the reach of inclusion. To expect every person to suddenly become AI-adept and capable of generating income in this new landscape is unrealistic. History makes this plain. Every previous industrial revolution left significant portions of society economically displaced, not temporarily, but permanently. Dependency on social welfare increased, pressure on the tax base intensified, and inequality widened. There is no credible reason to believe this revolution will be different, and considerable evidence to suggest it will move faster and cut deeper than those that preceded it. What is required is a focused, mandatory strategy that is driven and funded by those who profit most from AI systems and tools, to actively counter this outcome. The beneficiaries of AI advancement cannot be permitted to accumulate wealth on one side while societies absorb the human cost on the other. Redistribution mechanisms, and community-level economic transition plans must form part of any serious global AI governance framework. This is not a peripheral concern, it is foundational.

South Africa presents a cautionary but instructive case study in AI governance gaps. The country published its AI Policy Framework only in 2024, already significantly behind the pace of AI development, and received a mere 32 public submissions in response. This low level of engagement reflects a broader challenge: policy conversations that are inaccessible to ordinary citizens and civil society, result in governance process becoming the preserve of a narrow group, producing frameworks that lack legitimacy and democratic grounding. The gap between policy development and public participation became dramatically apparent in April 2026, when the Draft National AI Policy was published for public comment and withdrawn within three days. Investigative journalists discovered that the document contained fictitious and misrepresented references, sources that did not exist or were materially inaccurate. A national AI policy framework that cannot be trusted to accurately represent existing knowledge raises serious questions about the integrity of the process, the capacity of those responsible, and whether AI tools were used uncritically to generate the document itself, without appropriate human oversight or verification. This incident directly illustrates the governance gaps in my selected priorities. It is a transparency and accountability failure. It is a human oversight failure. And it is a consequence of insufficient capacity; not of the technology, but of the institutional structures meant to govern it.

• The Global Dialogue can play a transformative role in international cooperation but only if it moves deliberately beyond conversation toward binding commitment. The world does not lack opinions on AI governance. What it lacks is a trusted, neutral forum where those opinions are translated into internationally accepted standards that nations are genuinely accountable to. The Dialogue has the potential to be that forum, but it must resist becoming another high-level conversation that produces a communiqué and little else. Concretely, the Dialogue can advance cooperation in three ways. First, by establishing a shared definitional foundation. Nations cannot cooperate on AI governance if they cannot agree on what they are governing. The Dialogue should produce internationally accepted definitions of: AI-generated content, machine intellectual property
• and accountability thresholds. This will provide a common language for legislation across jurisdictions. Second, create enforceable minimum standards rather than aspirational guidelines. Voluntary frameworks have demonstrably failed to constrain the pace or ethics of AI development. The Dialogue should advance a model closer to international human rights law
• where minimum standards are agreed, adopted into domestic frameworks, and subject to review mechanisms. Third, by ensuring that the Global South has genuine influence, not token representation, in shaping these standards. South Africa's experience illustrates what happens when governance capacity does not keep pace with technological development. International cooperation must include structured support for nations building their AI governance frameworks from the ground up, ensuring that the rules of the road are not written exclusively by the nations and corporations with the most to gain from the current trajectory. The Dialogue's greatest contribution would be to make unaccountable AI development politically and legally costly.

The United Nations Educational, Scientific and Cultural Organisation's Recommendation on the Ethics of AI, adopted in 2021, represents the most broadly endorsed international normative framework to date. The Dialogue should treat this as its ethical baseline and focus its energy on what UNESCO's recommendation lacks, which are enforceability and implementation mechanisms. The African Union's Continental AI Strategy and the work of the African Observatory on Responsible AI are particularly relevant from a Global South perspective. These initiatives represent genuine regional ownership of AI governance and should be elevated within the Dialogue rather than marginalised by the louder voices of technologically dominant nations. The OECD AI Principles and the Global Partnership on AI have produced valuable technical and policy groundwork, primarily among developed nations. The Dialogue's added value here is bridging, i.e. connecting these frameworks to the realities of nations where governance capacity, digital infrastructure, and public awareness are still developing.

The Dialogue will only be as credible as the breadth and authenticity of the voices that shape it. Stakeholder contribution must therefore be deliberately designed. Governments must come prepared to commit. Their role is to translate Dialogue outcomes into domestic legislation and to be held accountable through transparent reporting mechanisms. A Dialogue that produces recommendations governments can quietly ignore has no value. The private sector, and specifically the corporations profiting most from AI development, must be present but not dominant. Their technical expertise is necessary; but their commercial interests must be named and managed. Participation should be conditional on disclosure of relevant financial interests and existing governance practices. Civil society, community organisations, and legal practitioners bring proximity to the people most affected by AI's consequences. Their contributions must be structured into the Dialogue at every level, not simply confined to a single panel or a public comment period. Academics and researchers provide the evidentiary foundation. Most importantly, ordinary people must have a genuine voice. This requires the Dialogue to operate simultaneously at multiple levels, i.e. formal plenary sessions for governments and institutions, and accessible public forums, in multiple languages, that reach citizens who will never attend a UN conference but whose lives will be shaped by its outcomes. In terms of format, the Dialogue requires a permanent secretariat, regional working groups, a public-facing transparency dashboard, and a binding review cycle. Governance is a continuous, accountable process.

The most underrepresented voices in global AI governance discussions are, predictably, those with the least power and the most to lose! People with disabilities are almost entirely absent from these conversations, despite being among the populations most directly affected; both by AI's potential to assist and by its demonstrated capacity to discriminate through biased systems and inaccessible design. Their inclusion requires more than accessible conference venues; it requires that accessibility is built into the design of participation itself. Older persons are similarly overlooked. As workplaces and public services accelerate toward AI-mediated interaction, they are least equipped to navigate that transition are rarely consulted about its terms. Children and young people deserve dedicated representation. They will inherit the AI systems being built today. Youth councils with genuine input into Dialogue outcomes would begin to address this. Informal and gig economy workers (e.g. the drivers, domestic workers, and piece-rate labourers) have no organised voice in AI governance, yet face the most immediate threat of economic displacement. Trade unions, community organisations, and civil society bodies must be formally empowered to represent them. Indigenous communities and rural populations are routinely excluded, yet their languages, cultural knowledge, and ways of life are both harvested by AI systems and threatened by them. Inclusion here requires engagement in local languages and on local terms.

The most innovative format the Dialogue could adopt is also the most obvious one it has yet to embrace, i.e. meeting people where they are, rather than expecting people to come to it. Traditional conference formats are designed for those already inside the room. They reward eloquence in dominant languages, familiarity with UN procedure, and the ability to travel. They consistently produce the same conversations between the same people. If the Dialogue is serious about meaningful engagement, the format must be ambitious. At the community level, structured public dialogues should be held in local languages. Trained facilitators, not AI experts, should lead these sessions, translating complex concepts into lived experience. For example: what does AI mean for your job, your children, your clinic, your language? These are the questions that generate honest answers. Simulated impact sessions, where participants experience realistic scenarios of AI-driven decision-making affecting employment, credit, healthcare, or legal outcomes, would make abstract governance questions personal and immediate. Open digital platforms, designed for low-bandwidth environments and available in multiple languages, would extend participation to those without the means or mobility to attend in person. Critically, these must be genuine input mechanisms with visible feedback loops and not just comment boxes that end up unread.

The European Union's AI Act, which came into force in 2024, is the most comprehensive binding legislative framework to date. Its risk-based classification system (distinguishing between unacceptable, high, limited, and minimal risk AI applications) provides a practical template for proportionate regulation. Its requirement that high-risk AI systems be transparent, auditable, and subject to human oversight directly mirrors the governance principles this Dialogue must advance globally. Rwanda presents an instructive African example. Its national AI policy explicitly incorporates digital inclusion and local language processing as governance priorities, recognising that effective AI governance must reflect the linguistic and cultural reality of its population. Canada's Algorithmic Impact Assessment tool, developed by the Treasury Board, requires government departments to evaluate the risks of automated decision-making before deployment. It is publicly available, open-source, and adaptable, which is exactly the kind of practical instrument the Dialogue should encourage all nations to adopt and contextualise. Finland's AI literacy programme demonstrates that public understanding of AI, the foundation of meaningful governance participation, can be scaled rapidly and affordably. This directly addresses the layperson inclusion deficit that undermines governance processes globally. The OECD AI Policy Observatory provides a comparative database of national AI strategies and legislation. Expanding its reach, ensuring Global South representation, and making it genuinely accessible in more languages would significantly strengthen international knowledge-sharing.