Artificial intelligence is often described as a global technology. But governments and organisations around the world are increasingly seeking greater control over how AI systems are developed, deployed, and governed. Investment is flowing into domestic computing capacity, local AI models, and regional technology systems, driven by concerns about security, resilience, and dependence on external providers. This change has brought growing attention to the concept of sovereign AI. Once largely confined to policy discussions and strategy documents, the term now appears regularly in government announcements, cloud industry initiatives, and technology reports. Sovereign AI concerns control: who governs the data, models, and infrastructure that underpin critical AI systems, and whether those capabilities remain subject to the laws and strategic choices of the organisations that rely on them. The idea features prominently in Bain & Company’s Technology Report 2025. The report situates sovereign AI within a broader trend of technological fragmentation, in which governments are reassessing their dependence on foreign technologies amid geopolitical tensions, export controls, and shifting global supply chains. This discussion represents an expansion of the earlier debate around data sovereignty. While data sovereignty focuses on where information is stored and which legal frameworks govern its use, sovereign AI raises a broader question: who controls the systems that transform data into intelligence? Many jurisdictions have already introduced requirements concerning the location and handling of sensitive data, including health records, financial information, and security-related information. These measures aim to ensure that critical data remains protected under appropriate legal and regulatory frameworks. However, controlling data does not necessarily mean controlling AI capabilities. A public institution could store citizen data domestically while relying on a foreign proprietary model to analyse documents, assess risks, or support administrative decisions. In such a case, the data may remain under national jurisdiction, but important parts of the analytical process remain dependent on external systems. This is where the concept of intelligence sovereignty becomes important. Sovereign AI extends questions of control beyond data storage to the models, algorithms, computing resources, and governance structures that determine how AI systems operate. Sovereign AI is generally understood as the ability of a country or organisation to develop, deploy, and govern AI systems using infrastructure, data, and models that remain subject to its own rules and priorities. This does not require complete technological independence. Few countries possess the resources to control every layer of the AI system. Instead, it reflects an effort to reduce critical dependencies and retain meaningful choices over strategically important capabilities.
The growing interest in sovereign AI is closely connected to geopolitical competition and changes in global technology. AI development depends on a complex network of semiconductors, cloud infrastructure, computing capacity, specialised talent, and advanced models. Many of these capabilities are concentrated among a relatively small number of countries and companies. The competition between the United States and China has highlighted the strategic importance of these dependencies. Export controls on advanced chips and restrictions on access to certain technologies have prompted governments to consider whether reliance on external providers could create vulnerabilities. In this context, sovereign AI addresses two related concerns. The first is whether access to essential AI capabilities could be disrupted by geopolitical decisions. The second is whether systems used in sensitive areas, such as defence, public administration, and critical infrastructure, should depend entirely on technologies controlled elsewhere. Technology companies frame this in similar terms: building national AI capabilities through infrastructure, data, workforce development, and industry partnerships, and maintaining control over the broader intelligence supply chain, including hardware, data, and algorithms. These approaches point toward a common idea: AI sovereignty depends on multiple layers of capability. Hosting a model locally may address some concerns, but control also depends on access to computing resources, technical expertise, and the ability to govern system evolution. At the same time, complete independence remains unrealistic for most countries. Advanced chips, frontier models, and specialised knowledge continue to rely on international supply chains and collaboration. Therefore, sovereign AI does not mean eliminating dependence entirely. It becomes more about deciding which dependencies are acceptable and which create unacceptable risks. Although sovereign AI has become a common policy objective, countries and regions pursue it for different reasons. China’s approach places strong emphasis on technological self-reliance and control across the AI system, including infrastructure, models, and industrial capacity. This reflects broader national strategies aimed at reducing dependence on external technologies deemed strategically important. The European Union has approached sovereign AI through a combination of regulation, industrial policy, and infrastructure investment. European initiatives emphasise trustworthy AI, regulatory alignment, and the development of domestic capabilities while remaining connected to global innovation networks. Programs such as InvestAI and partnerships supporting European industrial AI aim to expand computing capacity and strengthen regional competitiveness. The European Union has approached sovereign AI through a combination of regulation, industrial policy, and infrastructure investment. European initiatives emphasise trustworthy AI, regulatory alignment, and the development of domestic capabilities while remaining connected to global innovation networks. The InvestAI initiative allocates substantial funding toward AI “gigafactories,” large data centres built around significant GPU capacity, and industrial AI cloud projects, such as the partnership between Deutsche Telekom and NVIDIA aimed at European manufacturers. These are part of the same push to expand computing capacity and strengthen regional competitiveness. In the Gulf region, sovereign AI is closely linked with economic diversification and the development of new technology sectors. Investments in domestic data centres and Arabic-language AI models reflect both a desire for greater control over critical digital infrastructure and an opportunity to develop capabilities adapted to regional languages and markets. These approaches demonstrate that sovereign AI does not represent a single model. For some countries, the priority is security and resilience. For others, it is regulatory control, cultural representation, or economic development. The common challenge is balancing greater autonomy with continued participation in global technology networks. Open source models, international research, and cross-border expertise remain important drivers of AI progress. Even the most ambitious sovereign AI strategies operate within a highly interconnected environment.
The rise of sovereign AI creates new challenges for multinational organisations. As AI becomes embedded in business operations, companies may need to adapt not only their compliance strategies but also their technical architectures. A single AI workflow may require different implementations across markets depending on data governance requirements, infrastructure availability, and regulatory expectations. Organisations may need to manage multiple AI environments rather than rely on a single universal approach. However, not every application requires the same level of sovereignty. Sensitive use cases involving confidential information, critical infrastructure, or regulated services may require greater control over data, models, and deployment environments. Other applications may continue to benefit from global cloud platforms and shared AI services. This suggests that sovereignty should be viewed as a layered concept. Organisations can choose where greater control is necessary while maintaining flexibility elsewhere. The challenge is finding the right balance between control, efficiency, and innovation. For multinational companies, the most effective approach may be maintaining strategic options. Organisations need to understand where dependencies create genuine risks and where global platforms provide valuable advantages. Sovereign AI also raises broader questions for international governance. Countries differ significantly in their access to computing resources, infrastructure, investment, and technical expertise. As some governments build stronger domestic AI capabilities, others may struggle to participate equally in the emerging AI economy. Greater national control over AI may also complicate efforts to establish shared international standards. Different jurisdictions may prioritise different values, from regulatory safeguards and fundamental rights to innovation and competitiveness. These differences could shape how AI systems are developed, evaluated, and deployed. However, sovereignty does not necessarily mean isolation. A more practical understanding of sovereign AI focuses on the ability to make meaningful choices while remaining connected to international cooperation. AI development will continue to depend on global research networks, supply chains, and exchanges of expertise. Sovereign AI reflects a broader shift in how societies understand technology. Artificial intelligence is increasingly viewed as a strategic capability connected to economic competitiveness, public administration, and national resilience. In most cases, countries or organisations cannot control every part of the AI system, so the more important question is which capabilities require greater control and where external dependencies remain manageable. The future of AI will likely involve a complex combination of domestic capabilities, international partnerships, open technologies, and negotiated dependencies. Sovereignty will be measured more by the ability to make informed choices about where control matters most, rather than by complete independence. In that sense, the debate around sovereign AI offers an early glimpse into how artificial intelligence is becoming part of the language of global affairs. It is a discussion about technology, but also about resilience, influence, and the ability to shape decisions in an increasingly interconnected world. Author: Slobodan KovrlijaFrom data sovereignty to intelligence sovereignty
Geopolitics, fragmentation, and technology
Different regions, different priorities
Implications for multinational organisations
Sovereign AI and the future of AI governance
Sovereign AI as a question of strategic choice
