How Africa and Europe could create global AI governance rules

As AI systems rapidly spread across borders, the rules that govern them tend to come from a handful of powerful players. 2025 saw Africa moving from bystander to active participant in the global AI governance debates. During the 7th AU-EU summit in November 2025, European and African policymakers reaffirmed their commitment to enhancing digital infrastructure and promoting human-centric, trustworthy AI. However, their efforts stopped short of delivering what is now urgently needed: a mutually beneficial AI governance framework that promotes safe, transparent AI, protects human rights, and mitigates risks. Achieving this will require Africa to develop a unified, principled approach to AI governance, while the EU must stop being one-sided and engage more deeply with African priorities, youth and the private sector.

AI is on the EU-AU agenda 

At the 7th Africa-EU summit, the AU and EU committed to strengthening regional innovation ecosystems, including human-centric and trustworthy AI that respects intellectual property rights. However, the summit failed to provide a unified operational framework for digital cooperation or AI governance. The summit’s declaration sounds weak and largely aspirational. It also does not reference key AU AI and data governance frameworks and overlooks AI-specific risks such as algorithmic bias, labour displacement, and exploitative data practices. 

By comparison, the May 2025 Ministerial joint communiqué was more ambitious.  The document expressed a commitment to align AI governance frameworks based on shared values –and provided real direction for co-creation and a powerful joint leadership on the global stage. Both blocs will need to establish a concrete roadmap with milestones to develop and harmonise their respective AI regulatory approaches to translate that vision into practice.

A common ground for joint leadership

First, the EU and AU need to find common ground in AI governance despite different legal and policy contexts. The EU has pioneered the AI Act – the first in the world comprehensive legal framework for AI – to ensure  human-centric, trustworthy AI while protecting health, safety and fundamental rights. However, the November 2025 EU digital omnibus postponed large parts of its implementation. This raised doubts about the law’s immediate practicality and, by extension, its usefulness for external partners, like Africa, hoping to align with or learn from it. 

Still, the EU’s broader International Digital Strategy promotes a mutually beneficial approach to AI partnership. It acknowledges Africa's context and aligns efforts for responsible and human-centric AI uptake, focusing on areas like skills, infrastructure, and an enabling environment for innovation. In addition, the EU’s AI Continent Action Plan and Apply AI Strategy focus on capacity-building and practical use cases – in line with the AU's development priorities. 

On the AU’s side, the Continental AI Strategy provides a governance blueprint for member states, positioning AI governance as essential to managing risks, advancing inclusivity, human rights, gender equality, safety, and information integrity, while respecting African contexts, cultures and values.  

This convergence creates a strong basis for joint leadership: blending the EU's regulatory strength with the joint focus on development. While Africa may benefit from engaging with EU regulations, it should define its own AI governance agenda driven by domestic priorities and selectively adapt relevant elements to its context. As the EU itself reassesses its regulatory influence amid the evolving Brussels effect, both regions have the opportunity to co-create an AI governance framework which could serve as a powerful global reference and fulfil recent political commitments. 

Overcoming frictions on data governance and digital sovereignty

Secondly, effective co-creation on AI governance requires addressing past tensions, particularly around data governance and digital sovereignty. These frictions have surfaced when European companies operating in Africa prioritise GDPR compliance over domestic laws, effectively bypassing emerging African data protection frameworks and domestic legal authority. 

While the EU promotes cross-border data flows under strong data protection standards, many African states emphasise data sovereignty to retain local control over data for economic and security reasons. Increasingly, African policymakers frame ethical AI governance as central to building digital sovereignty, ensuring that regulatory frameworks reflect African values and support just and equitable growth. In contrast, European debates on digital sovereignty are more advanced and coordinated, focusing on EU autonomy and reduced dependence on non-EU companies.

To avoid repeating past mistakes, co-creation must move beyond a ‘standards-setter versus standards-taker’ dynamic. A joint framework should draw on African instruments like the AU Data Policy Framework, and incorporate equitable benefit-sharing and promote robust data governance that includes citizen and community participation, addresses structural inequalities, and tackles infrastructural, data, and skills deficits across the AI value chain. 

Building a unified vision

Thirdly, the two regions should establish a shared baseline of values and a unified vision. At the G20 summit, South Africa and the AU successfully advanced the African principle of ubuntu – often translated as “I am because we are” – into the G20 leaders’ declaration. Beyond a social value, ubuntu is seen by experts as a foundation for authentic innovation that prioritises solving collective human challenges rather than maximising profit. Integrating such ethical principles into a joint AU-EU framework would strengthen trust, ensure culturally grounded and socially responsive governance, and align closely with the EU’s commitment to inclusive, values-driven global cooperation under its International Digital Strategy. 

A joint framework should combine the EU's regulatory experience with the AU Continental AI Strategy and be grounded in the wider ethical framework set out in the UNESCO Recommendation on the Ethics of Artificial Intelligence. This multi-layered approach would ensure equality and foster fundamental rights. To this end, the EU-AU partnership should establish systemic principles, including: 

1. Preventing ‘digital colonialism’ by supporting locally developed AI models, infrastructure and African-centric datasets to reduce bias and external dependence.
2. Strengthening accountability by empowering institutions such as data protection authorities or AI offices through capacity building.
3. Conducting robust risk assessments, including Fundamental Rights Impact Assessments for high‑risk AI systems like biometric identification tools. For this, the EU’s AI Act is a useful model for both regions to limit AI use to what is legally necessary.

EU-Africa AI partnership in practice

Existing initiatives and actors like D4D Hub or other Team Europe Initiatives are crucial to building a joint EU-AU AI framework or new mechanisms to support research, innovation and funding for AI development in Africa. Beyond strengthening the African AI ecosystem and creating opportunities for EU private sector investment, the partnership should also incentivise African AI solutions to access EU markets. Regulatory sandboxes offer a practical way forward, providing safe environments to test African AI systems and demonstrate compliance with EU regulatory requirements.  

The AU-EU partnership is at a critical juncture, capable of defining the future of global AI governance. By connecting the EU's pioneering AI Act, the AU's Continental AI Strategy, alongside values like ubuntu and global frameworks, this partnership can co-build a human-centric model for AI governance. This shared vision must not forget systemic safeguards to mitigate AI risks, prevent human rights violations, and put an end to digital exploitation, while stimulating investments and innovation across both continents. 

The views are those of the authors and not necessarily those of ECDPM.