New paper from the TBI warns against AI isolationism in pursuit of sovereignty.
Authored by government technology advisors, the paper argues AI sovereignty is about strategic agency, not self-sufficiency.
The paper recommends a new “Control, Steer, Depend” framework for governments to identify and leverage AI strengths.
Tony Blair: “Countries that treat AI as a central pillar of their national purpose, deploying it widely and negotiating their place in the global ecosystem with clarity and ambition, will not see their sovereignty eroded.”
Pursuing AI isolationist policies to protect sovereignty would be a mistake for most countries, according to the Tony Blair Institute for Global Change. Sovereignty in the Age of AI: Strategic Choices, Structural Dependencies and the Long Game Ahead, published today by the former British Prime Minister’s Institute, argues that no country can be fully sovereign on AI – and that attempts to have complete control over its infrastructure risk weakening, rather than strengthening, national power.
The report is co-authored by Hilda Barasa, PeiChin Tay, Keegan McBride, Alexander Iosad and Jakob Mökander, who have advised governments across multiple continents, and features a foreword written by Tony Blair.
Frontier AI development is increasingly concentrated among a small number of firms and countries, with costs and energy demands beyond the reach of most states. However, as the technology grows in geopolitical and economic importance, failing to access and deploy AI presents a major threat to economic development, public services and global competitiveness.
Today’s report warns that responding to this reality with isolationist approaches, such as attempts to build fully sovereign AI stacks, from chips to frontier models, is too slow, too expensive and, for most countries, unachievable. More importantly, such strategies could cut countries off from the most advanced AI capabilities, undermining economic competitiveness, public-service delivery and national security.
The authors argue that sovereignty in the age of AI is instead about agency, not maintaining total control and independence. Every country can make strategic decisions about where to build domestic strength, where to shape markets and standards, and where to rely on trusted partners.
For example, Nordic countries, where it costs less to cool and therefore run data centres, have a strong bargaining chip for countries that lack this capacity but excel in other areas such as semiconductor production or research and development.
The paper calls for a more realistic and strategic approach that embraces managed interdependence while expanding national agility.
In his foreword to the paper, Tony Blair said:
“No state can dominate every layer of the AI stack. Leaders must make deliberate choices about where they want to build strength and influence. And by becoming indispensable in specific parts of the AI ecosystem – whether in data assets, specialised models, regulatory standards, energy capacity or talent pipelines – countries gain leverage across it, even if they do not control it all …
“Countries that treat AI as a central pillar of their national purpose, deploying it widely and negotiating their place in the global ecosystem with clarity and ambition, will not see their sovereignty eroded. They will renew it for a new age.”
To help governments navigate these choices, the paper proposes a new “Control, Steer, Depend” framework for exercising agency in an interdependent AI ecosystem.
The framework acknowledges that governments need direct control over certain critical systems, such as sensitive data or mission-critical public-sector applications, in addition to country-specific strengths such as research capabilities or clean energy that can be leveraged for export.
In some areas, they can steer outcomes through regulation, procurement, standards and partnerships. And where dependence on external providers is unavoidable – particularly for frontier capabilities – they can depend in ways that are negotiated, diversified and resilient, avoiding excessive concentration and vendor lock-in.
By deliberately and strategically balancing which parts of the stack they will control, steer and depend, nations can maintain their sovereignty and sphere of influence without building a fully sovereign AI ecosystem.
Dr Keegan McBride, Senior Policy Advisor for Emerging Technology and Geopolitics at TBI and one of the paper’s authors, said:
“No country can do this on their own: what is important is being able to act strategically – striking agreements and forging new tech-alliances.”