Strategies for sovereignty in the Age of AI

Details: James Barton; Category: Enterprise Ecosystems; 22 July 2026; 2105 views

As a topic, AI sovereignty has gained exponential traction in the last six months as enterprises have begun bringing AI use cases into operation – and enterprises worldwide, particularly in developing markets, will be keen to ensure that their long-term strategies are compliant with the various regulatory frameworks that are emerging worldwide.

The fear of overreliance on AI from another sovereign nation is now beyond theoretical debate – in June 2026, the USA suspended access to Anthropic's most capable models by export-control order, with no allied consultation, and rejected a UK request for exemption at the G7 summit.

Matt Walker of MTN Consulting told us: “that episode made clear that access to frontier AI is contingent on US political will, regardless of commercial relationships…any government relying on a US-based AI vendor needs to accept that this reliance is a serious liability.”

This is a major factor alerting governments and enterprises to the importance of AI sovereignty. Jan Wuppermann, Head of Service Assurance, Data and AI, NTT DATA, notes that the past few years have seen “a broader geopolitical decoupling of the world, where certain things are not taken for granted anymore, in terms of dependencies or lack of sovereignty.” Nations and enterprises are focused on what they have to do to stay in control: “Every nation wants to be a winner in the race for AI… a rule maker, and not a rule taker in this game.”

In the age of AI, model sovereignty is now top of the agenda – where do these sit, and how much control can governments or enterprises have over them? If the data is sovereign to a particular country but is fed by a model located elsewhere, what does this mean for both infrastructure and the models? Do they need to be brought onshore to ensure effective full stack sovereignty?

Self-sufficiency

Wuppermann notes that this is not just a national legal jurisdiction question – enterprises have realised that AI sovereignty isn’t as straightforward as they initially thought, which has provoked an opposite reaction towards more fragmented infrastructure and solutions. Against the geopolitical backdrop and supply chain disruptions, the idea that companies must become more self-sufficient along the value chain has been amplified. Walker argues that the unpredictable US president worsens the problem significantly, noting that while Trump won’t be in charge forever, the fact that his nativist, isolationist tendencies have gotten so much support in America is giving the rest of the world pause.

Wuppermann agrees that geopolitical trends rather than technology trends are driving the shift to a more national or regional fragmentation across the data centre side, jurisdictions, and models. “Enterprises do not want to be dependent on one single provider. They're starting to realise that if they put everything on one model - Chat GPT, Anthropic, Mistral – they create too much strategic dependency on it.” Instead, businesses are understanding the need for greater strategic interdependence as the complexity is too vast to manage otherwise; the challenge is to reconcile this with their own need for independence and sovereignty.

Sovereignty concern is close to universal across all markets, but sovereignty capacity is not. While AI sovereignty is being pursued unevenly across different jurisdictions, Walker argues that this divergence is driven by capital, power, and infrastructure constraints rather than a lack of concern.

“Developed economies and wealthy states possess the capital to build regional AI factories and subsidise domestic silicon manufacturing”, notes Walker, citing the examples of Abu Dhabi's MGX closing a $49 billion fund in July 2026 to buy across the entire AI stack, Saudi Arabia building domestic data centres, Qatar launching an infrastructure-only joint venture, and Singapore taking equity stakes in AI labs rather than building infrastructure. “These markets are all well-capitalised and able to pursue sovereign AI, although with different strategies. Poorer countries share the same anxiety but lack the capital, the chip access, or the domestic engineering talent to act on it.”

Given these disadvantages, Wuppermann expects emerging markets to work to their own timelines regarding AI sovereignty – and he highlights that shareholders and their representatives on the board must take accountability here, arguing that firms in developing countries are not adequately prepared for the balancing act between sovereignty, safety, and competitiveness. “There needs to be a significant wake-up call - you can't just wait for regulators to make your life easier. You’ve got to take accountability at the board level and therefore drive and navigate AI governance much more proactively at the enterprise level.”

Proactive Approach

Walker agrees that businesses in developing countries will need to take a proactive approach towards sovereignty to acquire and maintain control over the infrastructure, data and models that power AI. Full-stack control demands four layers working together: chips and compute, the fibre and satellite links connecting compute to users, the data flowing through it, and the models trained on it. “Most sovereign AI plans stop at the data center and call it done - that is a mistake. Owning a GPU cluster means nothing if a foreign entity controls the cable feeding it, or if communications running over that network can be intercepted by an adversary. Single-chokepoint control is not sovereignty. You need leverage across the stack.”

While full-stack sovereignty is of critical importance, it has brought challenges and costs - particularly for governments in lower-income countries - but Walker highlights that alternative approaches can allow businesses to avoid some of these difficulties. He cites the example of MTN’s proposed ‘data embassy’ which leverages the operator group’s footprint across Africa to host governments’ data and sovereign workloads in secure, extraterritorial facilities abroad. This bypasses the need to build expensive local data centres; since international operator groups maintain extensive network footprints across emerging markets, they can provide the connectivity required to link to these external nodes. Data centre capacity is sparse in regions like Sub-Saharan Africa, so this approach offers a practical mechanism for states to achieve data sovereignty without waiting for domestic infrastructure buildouts. “This is a strong fit for emerging markets specifically because licensed regional telcos like MTN Group are more common across Africa, Latin America and South Asia than hyperscale data centres are”, says Walker.

The Big Seven

While it has not yet been widely adopted, MTN’s data embassy model is an example of an innovative solution to the sovereignty issue born from local understanding of emerging markets – and it highlights how players in these regions may need to develop their own frameworks for AI sovereignty rather than passively adopting the ‘Big Seven’ legislations around AI emerging from the EU, UK, USA, Australia, India, China and Singapore. Wuppermann notes that while the intent of these frameworks is broadly the same - “They all want safety and innovation - they're all aspiring to find that right balance” – he acknowledges that from a compliance perspective, they are already hard to follow, running on different timelines and with varying implementation approaches.

The complexity of these different frameworks is obviously a hurdle, and Wuppermann believes that a coordinated global approach would be beneficial for all markets – however, there are a lot of issues surrounding this. Notably, selecting a responsible governing body would be a challenge – he argues that the current geopolitical climate somewhat undermines the OECD given that “the belief in global institutions to help govern certain things is weakened at the moment.” Instead, Wuppermann considers a regional cluster alignment as the more likely outcome: “I don't see a short-term global standardisation trend there. I wish there was a body like this, and I think that there should be. There should be more responsibility put outside of enterprises and the industries into a governing body - not meaning control, just more alignment and accountability.”