Skip to content

AI Glossary · Governance & compliance

What is Data Sovereignty?

Data Sovereignty

Data sovereignty is control over data and the workloads that process it: what data gets used, where it's stored, when it's processed, and which data center it runs in. Legally it's a question of jurisdiction, but enterprises increasingly use the word to mean operational control — the ability to move a workload, or pull it home, when they need to.

Also known as: data localization, sovereignty

· Updated · Chain of Thought

Enterprise AIAI Infrastructure

  • 2 episodes
  • 2 sources

The legal sense is about jurisdiction: which country’s laws govern the data. That isn’t only a question of where it’s stored, since the GDPR, for instance, also applies to some processing done outside the EU. The narrowest version is data localization, a rule that certain data must stay in a country, and it’s the part regulators write down most explicitly. Not every rule is a localization rule, though. The GDPR, for example, doesn’t require personal data to stay in the EU; its Chapter V allows transfers to other countries only under set conditions. Enterprises have started using “sovereignty” to mean something wider — the ability to choose where a workload runs, move it between locations, and pull it back into a local data center on demand. In episode 62 at 30:01, Kris Lovejoy says the word “actually just means control,” and that the control has to hold at the operational, technical and data layers, not just at the storage address.

For AI, Lovejoy puts the driver in supply-chain trust. At 38:35 she says many nation states fear technology being used as a tool of trade: imagine Amazon or Microsoft is suddenly no longer available. “Whether it’s credible or not credible,” she says, it is something people fear. On her account, that fear is doing real architectural work: it pushes teams toward AI infrastructure that can shift workloads across hybrid and multi-cloud environments rather than committing permanently to one provider or region, and it’s part of why open source keeps surfacing in the same conversation. On the training side, federated learning is one technique for learning from data held in several places without pooling the raw examples in one location.

Check every place the data goes

A storage region answers one question. An invented case: an AI service keeps a team’s documents in an approved region. The UK NCSC’s cloud security guidance asks for more than that. It wants the countries where data is stored, processed and managed, the legal jurisdictions it falls under, and the rights the provider has to access and use it. It also covers derived data such as verbose logs and machine learning models, and it notes that data stored in a local region can still depend on an authentication service replicated worldwide.

So list each flow separately: where documents are stored, where inference runs, where logs, backups and any fine-tuned model live, and from which countries support staff can reach them. Then ask the data and legal owners which requirement applies to each flow. A region setting is evidence about one of them, not a sovereignty guarantee.

Sources

  • GDPR Article 44: General principle for transfers — Allows transfers of personal data to a third country or international organisation only under the conditions of GDPR Chapter V, the legal core of keeping data under EU rules; checked October 26, 2026.
  • NCSC: Cloud security principle 2, asset protection and resilience — Asks customers to know the countries where data is stored, processed and managed, the jurisdictions it is subject to, and the provider's rights to access it, including derived data such as logs and machine learning models; checked November 7, 2026.

From the conversation