Towards a Fourth Digital Empire?

On 6 and 7 July 2026, representatives of UN Member States met in Geneva with civil society, academia, business and the technical community for the first United Nations Global Dialogue on AI Governance. The meeting was intended to broaden a discussion in which countries with less AI capacity have had limited ability to shape international debates that directly affect them. Participants came from across the world, but the vocabulary commonly used to describe digital power did not fully reflect the diversity of their perspectives and experiences.

This gap is also reflected in influential accounts of the global digital order. In Digital Empires (pp. 6-7), Anu Bradford identifies three centres of influence: a United States model led by private enterprise, a Chinese model combining state direction with industrial capacity, and a European model relying largely on regulation. The framework explains why choices made in Washington, Beijing or Brussels travel beyond their places of origin. It is less satisfactory as a complete map of where normative priorities emerge. Countries outside the three centres then appear mainly as recipients of European rules, American platforms or Chinese infrastructure.

The African Union’s (AU) Continental Artificial Intelligence Strategy offers a useful way to examine that limitation. It should not, however, be read in isolation, but as part of a broader continental framework that includes the African Union Convention on Cyber Security and Personal Data Protection, the AU Data Policy Framework and the Declaration of Principles on Freedom of Expression and Access to Information in Africa. The Strategy is considered as a recent and particularly relevant example of how shared priorities may be articulated and positions coordinated before international negotiations reach their final stages. The question is therefore not whether a single instrument amounts to evidence of a fourth digital empire, but how the African Union can exercise normative influence when participation is formally broad while control over infrastructure, agendas and concepts remains concentrated.

Looking Beyond Regulatory Alignment

Bradford’s framework rightly shows that legal influence travels through material relationships (p. 20). European legislation affects firms abroad because access to the internal market is valuable. American companies operate platforms and cloud services far beyond the United States, while Chinese financing and equipment have shaped telecommunications and public infrastructure. Procurement decisions, technical standards, commercial dependence and switching costs all help determine which rules acquire effects abroad.

Yet a focus on the three most powerful centres can reproduce the asymmetry it seeks to explain. Other jurisdictions become visible chiefly through their alignment with external models, while their governments, courts, regional institutions and civil society organisations receive less attention as possible sources of normative claims. Placing the entire Global South in a fourth category would not solve the problem. The label covers states with very different political institutions, regulatory traditions and economic positions. African states themselves have pursued heterogeneous approaches to data and AI governance, reflecting differences in regulatory choices, institutional settings and levels of digital development.

The Continental Artificial Intelligence Strategy does not seek to eliminate these differences. It identifies areas in which coordination may nevertheless be possible, including infrastructure, skills, research and innovation, public-sector applications, human rights and the distribution of AI’s benefits (pp. 3-5). Coordination need not mean uniform legislation. A regional document can establish priorities while leaving room for different institutional choices. Its significance lies in trying to ensure that African participation does not begin only after the available options have been formulated elsewhere.

What Regional Soft Law Can Contribute

Adopted by the AU Executive Council in July 2024, the Continental AI Strategy is a regional soft-law instrument that does not create directly enforceable obligations. It sets continental policy priorities and calls on Member States, the African Union Commission and other stakeholders to develop national governance frameworks, build institutional capacity and coordinate their approaches to AI.

Developed through an AU-led process involving multi-stakeholder consultations and cooperation among African and international institutions (pp. 16–17), the Strategy describes its approach as Africa-centred, development-oriented and inclusive. Its broad scope encompasses governance, infrastructure, computing capacity, skills, research and innovation, data, public-sector adoption and human rights (pp. 3-5). It therefore treats development and regulation as interconnected rather than sequential concerns. A public authority may regulate an AI service while depending on a foreign provider for the cloud infrastructure on which it operates; similarly, data-protection rules have limited practical reach where institutions lack sufficient storage, processing capacity or technical expertise. In such cases, the ability to inspect systems, negotiate contractual conditions and retain knowledge may matter as much as formal legislative authority.

The Strategy sits alongside the African Digital Compact, adopted at the same Executive Council session as Africa’s common position on digital transformation. The Compact places AI within a broader agenda encompassing digital infrastructure, data governance, inclusion, digital rights and Africa’s participation in global digital governance. Read together, the two instruments seek to connect technological capacity with a stronger African role in shaping international rules rather than treating capacity-building merely as a preliminary stage before governance begins.

An important institutional limitation remains, however: the Strategy does not establish a continental regulator. Kinfe Yilma and Kebene Wodajo therefore caution that AI strategies risk remaining insufficient unless they are translated into more robust governance arrangements (pp. 239-241). At the same time, several concepts used in the Strategy – including responsible AI, human rights, safety, data governance and risk-based regulation – also circulate widely in international and European AI governance frameworks, including the UNESCO Recommendation on the Ethics of Artificial Intelligence and the EU Artificial Intelligence Act. The relevant question is therefore not whether these concepts are uniquely African, but how they are interpreted and combined with priorities that the Strategy presents as specific to African contexts, including development, infrastructural capacity, cultural values and regional cooperation.

The significance of soft law lies precisely in these coordinating and agenda-setting functions. Regional instruments can provide governments and diplomatic missions with common reference points, inform subsequent legislation and procurement practices (p. 35), and structure positions in international negotiations. They may also contribute over time to broader processes of norm formation, although the emergence of customary international law would still require the distinct elements of sufficiently general state practice and opinio juris. The immediate significance of the Strategy lies less in creating legal obligations than in organising priorities and providing an institutional basis for coordinated action.

The AU’s 2024 Common African Position on international law in cyberspace offers a precedent for articulating shared positions in a complex field. The Continental AI Strategy could support similar coordination if Member States and regional institutions use it as a common reference point in domestic policymaking and international negotiations.

Rights, Infrastructure and Partnership

The relationship between material capacity and rights becomes especially clear when considering the representation of African languages in large AI models. Recent advances have improved support for some widely spoken languages, but coverage remains highly uneven and much of Africa’s linguistic diversity is still poorly represented in current language technologies (Hussen et al., 2026, pp. 4-5, 9). This is often treated primarily as a technical limitation. Yet once AI systems are used in public administration, education, health services or political communication, the implications become broader. Uneven linguistic representation may shape access to information and public services and may also affect the ability of communities to participate effectively in decisions that concern them.

The African human rights system offers resources for analysing these consequences, even though it does not provide a ready-made legal answer to every issue raised by AI. The African Charter on Human and Peoples’ Rights combines individual and peoples’ rights with self-determination, development and control over natural resources. In Resolution 473 (2021), the African Commission on Human and Peoples’ Rights committed to studying the human rights implications of AI, robotics and emerging technologies, while calling for governance frameworks attentive to African values and circumstances. These sources allow technological dependence to be considered not only in terms of privacy and consent, but also through questions of access to knowledge, communal interests, collective economic value and public authority over essential services.

The right to development does not mean that every technological imbalance constitutes a human rights violation, nor does the collective dimension of the African Charter provide a simple legal solution to technological dependence. It does, however, make distributive questions harder to leave outside the analysis. Computing capacity, language resources and technical expertise cannot be reduced to matters of development assistance when they directly affect whether states are able to regulate effectively and participate meaningfully in international rule-making.

These issues become particularly tangible in the partnerships through which the Continental AI Strategy may be implemented. In February 2026, the African Union Commission announced a Memorandum of Understanding with Google covering cloud and AI infrastructure, skills, research, entrepreneurship and responsible governance, presenting the initiative as part of an effort to strengthen Africa’s “sovereign digital capacity”. Such cooperation does not in itself undermine regulatory autonomy, but it is not politically or institutionally neutral either. Where critical infrastructure, cloud services, access to data and technical expertise remain concentrated in a small number of foreign firms, partnerships can both expand technological capacity and reinforce structural dependence.

This tension is central to critical discussions of digital colonialism in Africa (e.g. Okechukwu Effoduh, pp. 302-304), which draw attention to the ways in which control over data and infrastructure may reproduce asymmetries between African institutions and foreign technology companies. The concept should not be used to characterise every cross-border technological partnership as a form of colonial domination. It does, however, bring into view questions of ownership, control, value extraction and technological dependence that can disappear when cooperation is framed only in terms of capacity-building. A related concern appears in Reichman and Purnhagen’s discussion of “Data (De-)Colonialisation?” (pp. 956-957), which connects digital regulation to control over data and to the ability to resist external forms of data extraction.

The public announcement of the AU–Google partnership does not disclose the detailed legal and contractual arrangements governing its implementation. This leaves several important questions unresolved: who will control the resulting infrastructure; under what conditions data may be accessed and processed; whether systems will be interoperable; whether public authorities will be able to change providers without prohibitive switching costs; and how much technical knowledge and operational capacity will remain within African institutions. The key test of capacity-building should therefore not be the quantity of infrastructure or expertise made available at the outset, but the degree of autonomy that remains once the partnership matures or ends. Partnerships are more likely to strengthen sovereign capacity where technology and knowledge are transferable, systems remain interoperable, public institutions retain meaningful control over data and procurement decisions, and dependence on a single provider can be reduced over time.

Regional governance instruments can help translate these concerns into practice. Common procurement requirements, model contractual clauses, interoperability standards and safeguards concerning public data and critical infrastructure could give concrete effect to the Strategy’s broader commitments. On this account, capacity-building should ultimately be judged by what African institutions, regulators and researchers are able to control, develop and sustain independently, rather than simply by what external providers make available to them.

Participation Before the Final Draft

The Global Digital Compact and UN Global Dialogue on AI Governance promote inclusion, but meaningful participation requires a role in defining agendas and problems. The Continental AI Strategy could support this through advanced coordination and diplomatic capacity. Its effects depend on implementation, partnerships and material resources; institutions unable to inspect systems, retrieve data or change providers retain limited regulatory autonomy.

Within Africa’s broader legal and institutional framework, the Strategy also points to normative and epistemic influence: advancing priorities and concepts rooted in African legal traditions and experiences, as African scholarly contributions to the UN Global Dialogue illustrate. Such influence should be assessed alongside the market, infrastructural and regulatory power identified by Bradford (pp. 17–20), through both institutional outcomes and the framing of governance problems. The question is therefore not simply whether Africa is becoming a “fourth digital empire”, but whether its institutions can reshape global AI governance while navigating technological dependence, without being absorbed into the power structures described by Bradford.