The global race for AI supremacy is on.
The UK was the first to lead the charge, launching the AI Safety Summit three years ago at Bletchley Park, which united political and technology leaders. The US has repeatedly enforced its position of control with the largest players across the global market; however, China’s open-source dominance remains a constant concern to Western ambitions.
So-called ‘challenger’ countries are now throwing their hats in the ring. In February of this year, Sweden’s national AI plan set out the country’s ambition to become one of the world’s ‘Top 10’ nations for AI innovation.
However, one of the key aspects that is constantly being discussed in this international competition is sovereignty, both of technology and infrastructure.
The debate is not straightforward.
AI sovereignty means control and collaboration
When we think of AI sovereignty, we imagine tight, nation-led controls. We’re talking about based and built domestic AI infrastructure, model providers, workers, and regulations. Yet, meeting AI demands in practice means that the ‘keep things on home turf’ argument is devalued.
A simple fact is that AI relies on cloud infrastructure.
The UK, for example, does not have sovereign hyperscale provider capability, at least not to match the scale of Microsoft, Google, or AWS. Frankly, it’s too late and too expensive to try and challenge. Therefore, the ambition of stamping ‘Made in Great Britain’ across AI infrastructure, products, and services is practically impossible. It’s the same for most countries worldwide. Exceptions include US and China, which still lose some domestic control as they scale into new markets.
Keeping everything related to AI within borders is not only implausible, but it also ends a nation’s ability to perform on the global stage. This isn’t to say that home turf projects and infrastructure are not advantageous – they fuel domestic development and workforces – but it does caution the knee-jerk response of keeping projects in-country when it is more suitable (and can be similarly controlled) elsewhere.
If discussions of sovereignty around AI start and end with geographical borders, it has failed to consider the reality.
Control, collaboration, and security
Sovereignty is also a question of security. In areas such as sensitive government or defence workflows, domestic storage and strict access controls might be essential. However, it’s not always the case.
An obvious example is the experience of Ukraine, which has shown that in a conflict environment, it can be more secure to replicate critical data outside national borders than to keep everything in one place. A physical threat to infrastructure can quickly become a digital threat to national capability.
The real issue is informed control. When it comes to maintaining security, nations must understand where their data sits, how it is protected, how it can be recovered, and whether the architecture they have chosen is resilient.
In practice, sovereignty is more about control. With visibility and understanding, nations can trust cross-border collaboration.
Providers as infrastructure: the pitfalls of model-dependence
Even with informed control, there’s a common pitfall lurking: single provider lock-in.
The consensus is that opting for a single AI provider, whether a model, tool, or infrastructure reduces complexity and lowers cost on multiple fronts. The trouble is that reliance on a single provider means risking exposure to changes beyond customer control.
At the extreme end, a provider could be acquired by an entity that enforces different control protocols for customers, a provider could relocate to a less secure region or country, or change regulatory boundaries.
Reliance also means vulnerability to product and price changes. If service becomes implausible, offboarding all systems from a single provider would be time-intensive and costly.
Keeping ‘all your eggs in one basket’ affects the day-to-day. Dependence means a lack of flexibility to adapt workloads and tailor model strengths and weaknesses to new use cases, markets, or operations as technology evolves.
As AI scales, understanding that different models and tools are needed for different tasks saves significant cost drains while also improving outputs. These two crucial considerations are vital in a nation’s ability to use AI effectively and efficiently, and to compete at a global level.
Also, who’s to say the leading AI providers of today may not be the leading model providers in three years’ time?
Regardless of whether AI is deployed at a national or corporate level, a model-independent approach guarantees greater control.
Workforces and retaining talent
With so many regions competing for AI talent, workforces are critical to a nation’s ability to build sovereignty and compete internationally.
Currently, nations are scrambling to build talent bases, and only a few are excelling at creating a foundation of sovereign expertise that will support growth over the long term.
France is an example of building a sovereign workforce effectively, creating educational pathways that nurture domestic talent. Research states that, in 2024, 70% of France’s AI professionals had studied in the country. Germany has a different approach, with nearly half of its AI workforce educated at foreign institutions. In 2024, it also brought in 72% of Europe’s skilled immigrants through the EU Blue Card scheme, which specifically aims to attract high-level workers.
Regardless of method, nations must consider how to keep and build talent pools. Sovereignty and global competitiveness efforts will falter if workforces are emigrating to better opportunities. To aid retention, AI projects should remain domestic when it’s most impactful to do so, providing workforces with steady experience, support, and growth.
While shipping projects overseas may give the advantage to other nations, the balance lies in assessing when to keep operations domestic and when to leverage collaboration strategically.
In essence, AI sovereignty is not about isolation. The winners will not be the countries that build the highest walls around their data. They will be those who know how to protect their sovereignty while striking a balance between control and collaboration.
