The tech
Sovereign AI describes a country's or organisation's ability to keep meaningful control over its AI systems. It has several layers:
- Compute: where the chips and data centres are, and who controls them.
- Data: where information is stored and processed, who can access it, and which laws apply.
- Models: whether the organisation can access, adapt and keep using the model.
- Operations: who runs the system, who can change it, and who can switch it off.
Sovereignty is broader than storing data in-country. It's also about operational control, skills and dependencies. Open weight models can support it, but having the weights doesn't on its own make a system sovereign.
Why it matters
- Governance. Institutions can design systems around their own data rules and public responsibilities. What data can leave a country depends on the jurisdiction and the type of data, so check the rules that apply rather than assuming.
- Resilience. More control reduces the risk of a supplier withdrawing access, changing its terms or retiring a model. Dependence on imported chips, software and energy remains.
- Local capability. Running and adapting systems builds domestic expertise and keeps priorities local. That needs sustained investment in infrastructure, people and maintenance, not a one-off purchase.
In the real world
- India's IndiaAI Mission, approved in March 2024 with a budget of about ₹10,372 crore over five years, tackles several layers at once. It subsidises access to tens of thousands of GPUs for researchers and startups, shares Indian datasets through its AIKosh platform, and funds foundation models trained on Indian languages.
- At the Global AI Summit on Africa in Kigali in April 2025, 49 African countries, the African Union and Smart Africa endorsed the Africa Declaration on Artificial Intelligence. It commits to building African talent, data, infrastructure and markets, and announced a proposed $60 billion Africa AI Fund and an Africa AI Council.