Who Owns GeoAI? Justice, Data Sovereignty, and Climate Decision-Making in Africa- Ruvimbo Doreen Supiya

This policy brief explores questions of data sovereignty, epistemic justice, and inclusion in GeoAI systems shaping climate decision-making in Africa. It was developed as part of my research on AI governance and geospatial systems.1. 1. 1. Introduction: Beyond Technical Innovation

Artificial intelligence (AI) and Earth observation technologies are rapidly transforming how decisions are made in climate-sensitive sectors such as agriculture. In Africa, these systems hold immense promise, enabling drought monitoring, crop assessment, and environmental forecasting at unprecedented scales.

Yet, beneath this promise lies a deeper question that is often overlooked:

Who owns the data, the models, and the infrastructure that shape these decisions?

Across many African contexts, AI systems are developed using externally controlled datasets, processed through foreign infrastructures, and deployed without meaningful local participation. This creates a paradox: while AI is used to inform decisions affecting African communities, those communities often have limited agency in defining the problems, shaping the models, or benefiting from the outcomes.

2. The Hidden Problem: Structural Injustice and Epistemic Harm

Current discussions on AI in Africa often focus on access, innovation, and capacity building. While these are important, they risk obscuring a more fundamental issue: structural injustice embedded within AI systems.

AI systems do not exist in isolation. They are built on:

• Historical data inequalities

• Unequal access to infrastructure

• Power asymmetries in knowledge production

In geospatial AI (GeoAI), these issues are particularly pronounced. Satellite data, mapping platforms, and machine learning models are often controlled by actors outside the regions they represent. As a result, African environments are observed, modelled, and interpreted externally, with limited local influence over how knowledge is generated and applied.

This reflects what can be understood as epistemic harm, where knowledge systems, languages, and lived realities are systematically excluded from AI design and deployment. In African contexts, this manifests in the absence of local languages, indigenous knowledge systems, and community-defined priorities in both datasets and model design. AI systems do not simply overlook African realities — they actively reshape them through incomplete and externally defined representations.

This results in technical advancement accompanied by epistemic harm, where systems may be accurate, but are not necessarily fair, representative, or accountable.

3. Data Sovereignty and the Question of Ownership

At the core of this issue is data sovereignty, the right of countries and communities to control how their data is collected, processed, and used.

In the context of GeoAI, this extends beyond data to include:

• Ownership of AI models

• Control over data infrastructures

• Authority over decision-making processes

Without data sovereignty, African policymakers remain dependent on external systems to interpret their own environments. This dependency limits the ability to design context-specific policies for climate resilience, agriculture, and resource management.

Without deliberate intervention, AI systems risk becoming a form of managed extraction, where data is sourced from African environments, processed externally, and reintegrated as insights without equitable control, ownership, or benefit-sharing.

This raises critical ethical questions:

• Who benefits from AI-generated insights?

• Who is accountable when systems fail?

• Whose knowledge is included — and whose is excluded?

4. Inclusion Beyond Representation

Inclusion in AI is often framed as representation in datasets or participation in technical development. However, this definition is insufficient.

True inclusion requires:

• Participation in defining problems

• Involvement in designing systems

• Ownership of data and models

• Influence over how AI is governed

This raises a critical question:Where is the African in AI systems that are increasingly used to make decisions about African lives and landscapes?

African languages, knowledge systems, and lived experiences existed long before the emergence of modern AI. Yet, they remain marginal in many contemporary systems. This absence is not accidental, it reflects deeper structural inequalities in whose knowledge is valued and whose is ignored.

In this sense, absence in data is not a technical gap, it is a structural condition.

5. From Policy Frameworks to Execution

Many African countries have begun developing national AI strategies and digital policies. However, a key challenge persists:

The gap is not ideas — it is execution.

Policy frameworks often emphasise regulation, ethics, and high-level principles, but lack:

• Mechanisms for implementation

• Investment in infrastructure

• Integration with real decision-making systems

• Accountability structures

As highlighted in contemporary AI governance discussions, Africa does not lack policy frameworks — it lacks execution capacity, infrastructure, and ownership in system design.

To move forward, there is a need to shift from:

• Regulation → Design

• Frameworks → Deployment

• Principles → Practice

This includes piloting AI systems that are locally grounded, transparent, and aligned with real policy needs.

6. Toward Just GeoAI Systems

Addressing these challenges requires rethinking GeoAI as not just a technical system, but a socio-technical and political system.

A just approach to GeoAI should prioritise:

1. Data Sovereignty

• Local control over geospatial data pipelines

• Investment in African data infrastructure

2. Participatory Approaches

• Inclusion of communities in data collection and validation

• Integration of local and indigenous knowledge

3. Transparency and Accountability

• Explainable AI systems

• Clear responsibility in AI-driven decisions

4. Capacity and Ownership

• Strengthening local expertise in GeoAI

• Supporting African-led research and innovation

7. Conclusion: Reclaiming Agency in AI Systems

GeoAI has the potential to transform climate and agricultural decision-making in Africa. However, without addressing issues of ownership, inclusion, and governance, it risks reinforcing the very inequalities it seeks to solve.

The challenge is not only to build better models, but to ask:

Who defines the problem?Who builds the system?Who controls the data?And who benefits from the outcome?

A just approach to AI requires more than technical excellence. It requires confronting epistemic harm, redistributing power, and ensuring that those most affected by AI systems are not only represented, but actively shape them.

For Africa, this means moving beyond participation toward ownership — of data, models, and the futures they define.