Sovereignty was high on the agenda at the India AI Impact Summit in New Delhi in February. Many discussions focused on the capacity of the Global South to build and deploy AI models using domestic data and compute infrastructure, reducing overdependence on external actors.

On the opening day of the summit Michael Kratsios, Director of the White House Office of Science & Technology Policy, had this to say:

“Real AI sovereignty means owning and using best-in-class technology for the benefit of your people, and charting your national destiny in the midst of global transformations. It does not mean waiting to participate in an AI-enabled global market until you have tried and failed to build full self-sufficiency.”

An altogether novel definition of sovereignty — one in which someone else owns everything.

The Thiel protégé

Kratsios, despite graduating with a degree in political science focusing on ancient Greek democracy, is one of Donald Trump’s key advisors. In 2017 at the age of 31 he was appointed Chief Technology Officer of the United States, having spent seven years working as protégé, CFO and principal with Peter Thiel. Trump later appointed him third highest ranking official in the Department of Defense, where his responsibilities included oversight of DARPA — an appointment that raised eyebrows, with suggestions he was acting as a proxy for Thiel.

His worldview is not subtle. After providing testimony to the UN Security Council in September last year, he tweeted:

“The US totally rejects all efforts by international bodies to assert centralized control & global governance of AI. Ideological fixations on social equity, climate catastrophism, & so-called existential risk are dangers to progress & obstacles to responsibly harnessing this tech.”

The stack

Three days after inauguration, Trump signed Executive Order 14179 — Removing Barriers to American Leadership in Artificial Intelligence. It declared: “It is the policy of the United States to sustain and enhance America’s global AI dominance.”

Six months later, the White House published Winning the Race: America’s AI Action Plan, built on three pillars: accelerating AI innovation, building American AI infrastructure, and leading in international AI diplomacy and security.

The day before, Executive Order 14320 was signed — Promoting the Export of the American AI Technology Stack. Its purpose: “ensure that American AI technologies, standards, and governance models are adopted worldwide.”

The resulting American AI Exports Program calls for Big Tech consortia to create proposals for full-stack technology packages encompassing chips, servers, cloud services, data pipelines, AI models, security systems and specific applications. As the Institute for AI Policy and Strategy translated it:

“A consortium might look like a US model developer, such as OpenAI, working with a US cloud provider, such as Microsoft, and a chip company such as Nvidia, to build a data centre in a particular country to sell access to models for a particular set of applications.”

Note that the consortium doesn’t sell. It sells access.

Standards as governance architecture

Alongside the export programme, NIST launched the AI Agent Standards Initiative in February — a push to write the rules for agentic AI. NIST’s own language is refreshingly candid: the initiative exists to “foster the emerging ecosystem of industry-led AI standards and protocols while cementing U.S. dominance at the technological frontier.”

If the export programme puts American infrastructure under the bonnet, the standards initiative writes the rules of the road. Countries that adopt the stack will, by default, adopt the governance architecture — without necessarily having been in the room when it was designed.

There is also a National Champions Initiative which, in Kratsios’ words, “is designed to include the leading technology companies of partner countries directly into the American AI stack.” A champion may develop local applications built off the stack, but the stack itself remains firmly in US ownership.

Who pays, and why

As to financing, Kratsios spelled this out in New Delhi: the US Development Finance Corporation, the Export-Import Bank, the US Trade and Development Agency, the Millennium Challenge Corporation and a new World Bank Fund have all initiated AI-focused programmes.

Which raises the obvious question: if hyperscalers are planning to invest $680 billion in AI capex this year, why do they need government help? Because they won’t invest where there’s no demand. The programme exists to bridge the gap between commercial logic — which points hyperscalers to wealthy markets — and geopolitical logic, which needs American presence in strategically important but commercially marginal countries. This is the government subsidising market entry the private sector would not otherwise pursue.

What this means for Africa

The implications are profound. African countries that adopt the American AI stack will find themselves locked into a governance architecture designed in Washington. Their data will flow through American-owned infrastructure, governed by American-written standards, with sovereignty redefined as “the right to use what we’ve built.”

The alternative — building African-owned infrastructure, developing African standards, insisting on genuine data sovereignty — is harder, slower and more expensive. But it is the only path that leads somewhere other than a new form of digital dependency.

This is not primarily a technology question. It is a question of political will.

Bill Anderson is a data governance consultant and founder of Data Landscapers Ltd.