For decades, geopolitics was shaped by control over territory, natural resources, maritime routes, and military capability. In the twenty-first century, those factors remain decisive, but a new dimension of power has emerged above them: the capacity to develop, train, regulate, and govern artificial intelligence.
The issue is no longer simply who builds the most advanced models or manufactures the most sophisticated semiconductors. The competition also revolves around who will write the rules that will govern the operation of this technology for the rest of the world.
With that objective, China unveiled an Action Plan on International AI Ethical Governance on July 17, 2026, during the World Artificial Intelligence Conference (WAIC) held in Shanghai. The document was prepared under the coordination of the Ministry of Industry and Information Technology and is explicitly framed within the United Nations Pact for the Future and the Global Digital Compact.
This decision is far from incidental. While much of the international debate continues to focus on technological competition between the United States and China, Beijing seeks to place another issue at the center of the global agenda: the construction of an international architecture for governing artificial intelligence before its development permanently outpaces the regulatory capacity of States.
The distinction is profound. From China’s perspective, artificial intelligence is not merely a strategic industry or a market of immense economic value. It is a civilizational infrastructure destined to transform production, education, medicine, scientific research, public administration, finance, transportation, security, and virtually every sphere of human activity. For this very reason, Beijing argues that its development cannot be determined solely by commercial interests or corporate competition, but instead requires governance capable of safeguarding the global public good.
From this perspective, ethics ceases to be a collection of abstract principles and becomes a political technology. Governing artificial intelligence ethically means deciding how it will be designed, who will bear responsibility when harm occurs, which risks will be considered acceptable, and who will participate in defining those rules.
One of the central concepts of the plan is the establishment of governance throughout the entire life cycle of artificial intelligence. Although the expression may appear highly technical, it represents a significant departure from traditional regulatory approaches. Rather than supervising only finished products, the proposal calls for oversight mechanisms beginning at the very origin of every system. This encompasses data acquisition, training processes, model design, safety testing, deployment, subsequent updates, and even the eventual withdrawal of systems whose risks can no longer be effectively managed.
In practice, this means that responsibility no longer rests exclusively with the end user. Developers, data providers, technology companies, research institutions, and regulatory authorities all share responsibilities throughout the technological development process. Artificial intelligence therefore ceases to be viewed merely as a commercial product and instead becomes a comprehensive chain of shared accountability.
Another fundamental pillar is the classification of risks according to categories and levels. The underlying logic is straightforward: not every artificial intelligence system presents the same potential for harm. A model designed to translate documents does not pose the same risks as one capable of operating critical energy infrastructure, financial systems, medical diagnostics, or military decision-making. Consequently, the plan proposes regulatory obligations proportional to the potential impact of each application, avoiding both insufficient oversight and sweeping prohibitions that could hinder innovation.
The document also introduces the concept of agile governance. China recognizes that technological development advances far more rapidly than traditional legislative processes. A law may require years of parliamentary debate, whereas an artificial intelligence model can evolve within months—or even weeks. In response, the plan advocates regulatory mechanisms capable of continuous adaptation through technical standards, ongoing evaluation, and institutional coordination. Regulation is therefore conceived not as a static legal text, but as a dynamic and evolving process.
Another particularly significant aspect is the strengthening of a collaborative ecosystem. Under the Chinese approach, artificial intelligence is not viewed as the exclusive domain of major technology corporations. Universities, public laboratories, research institutes, manufacturing industries, open-source developers, international organizations, and governments are all considered integral parts of a single innovation chain. This reflects a longstanding characteristic of China’s development model, in which state planning seeks to coordinate capacities across multiple sectors in order to accelerate technological innovation.
Within this context, particular importance is attached to research on explainability, privacy protection, and bias mitigation. Explainability seeks to address one of the most complex questions in contemporary artificial intelligence: is it possible to understand how a model reached a particular conclusion? As AI systems become increasingly sophisticated, many of their decisions function as genuine “black boxes.” The plan therefore identifies as a priority the development of technologies capable of making these internal processes understandable while simultaneously strengthening personal data protection and reducing algorithmic discrimination.
Equally noteworthy is the explicit support for technological exchange and open-source development in areas related to security, transparency, and explainability. At a time marked by trade restrictions, export controls, and intensifying technological competition, this position seeks to project an image of international scientific cooperation while enabling countries with more limited technological capabilities to participate in the development of advanced AI tools.
The plan also incorporates a social dimension extending well beyond engineering. It proposes integrating scientific and technological ethics into the national education system, specifically safeguarding the rights and interests of women, children, older adults, and persons with disabilities, while reducing the digital divide. From China’s perspective, artificial intelligence should not deepen existing inequalities but rather serve as an instrument for expanding human development opportunities.
Perhaps the most significant element from a geopolitical perspective is the plan’s insistence on strengthening cooperation with developing countries. The document proposes expanding regulatory capacities, sharing institutional experience, and facilitating access to technical knowledge so that AI governance does not become concentrated within a small number of advanced economies. This orientation is closely connected to other initiatives promoted by China over the past decade, including the Belt and Road Initiative, the Global Development Initiative, and expanding technological cooperation with Asia, Africa, and Latin America.
This approach contrasts with the trajectory followed by the United States. Although Washington has also developed regulatory initiatives and recognizes the need to address risks associated with artificial intelligence, it has historically assigned a predominant role to private-sector dynamism and the preservation of technological leadership. Major corporations have played a central role in developing the world’s most advanced models, while public policy has combined regulation, innovation incentives, and measures designed to protect strategic capabilities, including export controls on high-performance semiconductors.
China advances a different approach. It argues that international governance should not be built solely around technological competition, but rather through multilateral mechanisms capable of establishing shared rules for the future development of artificial intelligence. From this perspective, regulation is not viewed as an obstacle to innovation but as a necessary condition for ensuring its long-term sustainability and for distributing its benefits more broadly.
Ultimately, the debate extends far beyond technology itself. What is now beginning to take shape is the normative architecture of the twenty-first century. The rules established today will determine who controls data, who sets international standards, who bears responsibility for technological risks, and who participates in the distribution of the benefits generated by artificial intelligence.
The plan presented in Shanghai reflects China’s determination to play an active role in shaping that architecture. Rather than merely responding to immediate technological competition, it seeks to occupy a central position in defining the rules of the emerging international digital order. The debate is no longer simply about who develops the most powerful artificial intelligence. The decisive question is who will possess the authority to define the rules under which that intelligence will ultimately transform the world.
Sources
- Ministry of Industry and Information Technology of the People’s Republic of China (MIIT). Official statements on the 2026 World Artificial Intelligence Conference.
- Xinhua News Agency. “Action Plan on International AI Ethical Governance,” July 17, 2026.
- United Nations. Pact for the Future and Global Digital Compact, 2024.
- Ministry of Foreign Affairs of the People’s Republic of China. Global AI Governance Action Plan, 2025.
- World Artificial Intelligence Conference (WAIC) 2026. Official speeches and conference documents.
In the boardrooms of Addis Ababa and the tech hubs of Nairobi, a new worldmaking project is underway. Artificial Intelligence (AI) has become one of the central arenas where African states negotiate their position in an unequal global order. To speak of “AI in Africa” is not simply to describe the arrival of a technology, but to encounter a set of competing discursive frames through which political desires, developmental ambitions, ethical claims, and governance priorities are translated into technical agendas. As national strategies and regional protocols are rolled out, what comes into view is a struggle over how Africa is imagined within this machine age, and what kinds of futures that imagination makes thinkable.
These frames are not simply different ways of describing a new technology. They are competing scripts for what Africa is, what AI is for, and who ought to shape its future. Some cast the continent as a site of leapfrogging and renewal; others as a zone of exposure, extraction, sovereignty, moral repair, or improvised adaptation. Each brings some actors and possibilities into focus while pushing others to the margins. That is why the issue is not only what AI means for Africa, but what pictures of Africa are being called into view through AI discourse.
The first terrain is shaped by promise and protection. One of the most pervasive registers is promissory. In the African renaissance 2.0 frame, visible in the AU Continental AI Strategy, AI appears as a leapfrogging device through which Africa can bypass the chokepoint of traditional industrialization and enter prosperity through technical modernization. Here, the ethical subject is the developmental state or planner charged with delivering growth, inclusion, and renewal. What comes into view is capacity, competitiveness, and beneficial transformation. What recedes from view are the historical and political-economic conditions that make acceleration so attractive in the first place: debt, infrastructural dependence, uneven ownership, and the possibility that AI may deepen rather than overcome inherited asymmetries.
Running alongside this is a protective register. In this vulnerability frame, Africa is not late but exposed to surveillance, bias, exclusion, and systems introduced into settings where safeguards are weak and redress is limited. As the CIVICUS report on AI surveillance in Africa suggests, the issue is not only future risk but already existing forms of technologically mediated harm. The ethical subject here is the harm-bearing population—the data subject, precarious worker, surveilled citizen, and linguistically marginal user most likely to absorb the costs of badly governed AI. This frame is ethically indispensable because it foregrounds rights, harm, and institutional fragility, including the uneven but growing architecture of data protection on the continent. Its limit, however, is that it can cast Africa chiefly as a site to be protected rather than one from which technological futures are actively contested and shaped.
A second terrain is shaped by structure and sovereignty. Alongside these promissory and protective registers sits a more structural critique. In the anti-colonial frame, AI appears as a socio-technical order built through unequal material relations: not only data extraction, but also mineral supply chains, outsourced labor, environmental degradation, foreign-owned infrastructures, and asymmetrical control over the institutions that govern digital life. From this perspective, Africa is not outside the AI order looking in; it is already deeply inserted into it, though often on subordinated terms. The ethical subject is the dispossessed worker, community, or polity whose labor, resources, and data are drawn into AI systems while ownership and value are captured elsewhere. What becomes visible is the political economy of AI—an approach that makes it harder to sustain the comforting idea that Africa’s main problem is simply one of readiness or delayed entry. But the frame has limits. In foregrounding dispossession, it can flatten internal differentiation, externalize agency too completely, and become so totalizing that the uneven realities of negotiation and strategic inhabitation recede from view.
The sovereignty frame shifts the question from extraction to control: AI is cast as a problem of authority over infrastructure, procurement, standards, and rulemaking, a concern visible in South Africa’s National Data and Cloud Computing Policy. Its appeal lies in insisting that African states must govern technological systems in the public interest rather than merely consume them on external terms; the ethical subject is the sovereign polity, imagined above all as the state charged with authoring technological futures on its own terms. But sovereignty should not be romanticized. Declaring authority is not the same as possessing material control, especially where states remain dependent on foreign cloud infrastructures, imported models, consultants, and opaque procurement chains. Equally, this frame can too quickly presume the state to be the vehicle of emancipation.
A third terrain is shaped by ethics and language. Beneath these macro-struggles lie cultural and ethical frames. In the Ubuntu frame, Africa appears as a moral resource: a source of relational vocabularies of dignity, community, interdependence, and solidarity that unsettle liberal individualism in mainstream AI ethics. At its strongest, this frame asks what kind of human subject AI presupposes and produces. Its ethical subject is the person-in-community, understood less as an isolated rights-bearer than as a being constituted through relations of community, solidarity, and mutual recognition. At its weakest, it risks remaining symbolically reparative and romantic rather than politically disruptive: a moral vocabulary that pluralizes AI ethics without redistributing authority over infrastructure, capital, or standards.
The linguistic justice frame relocates the issue to language, depicting Africa as linguistically marginalized. Initiatives such as Lelapa and Masakhane insist that language is not a neutral interface but a site through which recognition and opportunity are distributed in digital life. An AI system that cannot adequately process Yoruba, Wolof, or Zulu does more than inconvenience its speakers; it helps reproduce a hierarchy in which some languages, and the worlds they carry, are made computationally peripheral. The ethical subject here is the speaker, the language community, and the knower whose access to digital life is mediated through systems not built with them in mind. Yet linguistic inclusion alone does not resolve ownership, infrastructure, or power. An AI that speaks African languages can still be embedded in extractive or externally governed systems.
A final terrain is shaped by markets and adaptation. The market pragmatist frame, visible in the African Continental Free Trade Area (AfCFTA) Protocol on Digital Trade, casts Africa as a jurisdictional space to be made legible to capital—a continent whose frictions must be reduced so that innovation, trade, and platform expansion can move more easily across borders through harmonization, regulatory alignment, and investment readiness. Here, the ethical subject is the investor, firm, regulator, and entrepreneurial ecosystem said to require clear rules and scalable markets. What becomes visible is efficiency, interoperability, and growth. What recedes from view is the fact that harmonization is never merely technical. It can distribute advantage, privileging certain actors over others, and can also displace questions of labor and ownership while eclipsing democratic accountability behind the smoother language of market access.
The frugal innovator frame—the jua kali of tech—foregrounds small-scale, improvised solutions that work offline, on cheap devices, and in low-resource settings. Lelapa AI’s InkubaLM is the sharpest contemporary example. This frame’s appeal lies in refusing the assumption that meaningful technical work must begin from abundance. Africa appears as inventive under pressure, capable of repurposing constraints into forms of practical intelligence. The ethical subject is the builder, bricoleur, local entrepreneur, and community improviser who makes systems work where industrial-scale infrastructures are absent. What becomes apparent is creativity, resilience, and situated responsiveness. But it can also romanticize necessity, making scarcity look like authenticity rather than a political condition to be transformed. It can overstate what ingenuity can compensate for and understate the need for public investment, industrial policy, and durable institutional support.
In the end, what matters about these eight frames is not whether any one of them finally tells the truth about Africa and AI. It is that each organizes moral and political attention differently—foregrounding development while muting dependency, naming extraction and harm while narrowing agency, valorizing language and ethics without redistributing material authority. As a constellation of concerns, they reveal less a single African AI trajectory than a struggle over what Africa is imagined and allowed to be in technological modernity.
That is why AI in Africa is not just about chips, code, or competitiveness. It is about who gets to architect the future, under what terms, and in whose interests. The stakes are not simply adoption or regulation, but the representational politics through which certain subjects, institutions, and futures become thinkable while others fall from view.
Yet these frames do not only describe Africa; they help make it where they are taken up. This is the worldmaking with which we began. What a frame brings into view becomes what institutions are built to serve, and what it lets recede goes ignored. Nor do the frames work alone. They overlap and prop one another up: sovereignty talk can leverage the anti-colonial critique, market pragmatism needs the renaissance’s promissory optimism, and the frugal innovator’s ingenuity is enlisted to soften the very scarcity it works within. The future, then, will not be settled by one script defeating the others, but by which come to dominate the mixture—and whose interests that mixture ends up serving—as it hardens into the continent’s legal, institutional, and material infrastructure. Whether this moment marks a structural shift or the repetition of old patterns turns on that.
No comments:
Post a Comment