Services

Sovereign and institution-scale AI

National and institution-scale AI strategy that controls the dependencies that matter, compute, data, models, and talent, instead of the ones that do not. Sovereignty as leverage, not slogans.

The problem it solves

"Sovereign AI" is used to mean everything and therefore nothing, a national data center, a home-grown model, a slogan about independence. The real question is narrower and harder: which dependencies, if someone else controlled them, would control you? Compute, data, models, and talent are not equally strategic, and trying to own all of them is how sovereign AI budgets are wasted. This work replaces the slogan with a decision: what to own, what to partner, and what to buy, so a government or institution holds the few dependencies that actually matter and stops paying to control the ones that don't.

Who it is for

Governments, ministries, sovereign entities, and institutions large enough that their AI choices are strategic, where the wrong dependency is a national exposure, not a procurement footnote.

What you get

• A sovereign AI operating model matched to your real constraints
• A dependency and control map: what to own, what to partner, what to buy
• National or institutional AI strategy
• Governance and foresight wired into how decisions get made

How it works

1. Fix the sovereign objective, past the buzzwords
2. Map the dependencies that actually confer control
3. Design the operating model
4. Wire in governance and a foresight rhythm

Grounded in the region's real rules

Sovereignty is local. The GCC has no single AI statute, so data-protection law and sector regulators carry the weight; Africa is building governance under the African Union's Continental AI Strategy; Europe enforces the EU AI Act. A sovereign operating model has to fit the jurisdiction it lives in, which is why this work is scored against the Sovereign AI Governance Readiness Framework and mapped to each market in the regional guidance. Proven in practice: a Gulf ministry innovation model whose Chief Design Officer function was adopted as a government-wide standard.

Questions

Common questions

What does a real sovereign AI strategy look like for an institution without a sovereign fund's balance sheet?

It looks like choosing where you need control and accepting dependence everywhere else. Compute, models, data, and talent are four separate decisions, and very few institutions need to own all four. Most need control over their data and over the applications closest to their mandate, and can rent the rest. A workable strategy names those boundaries, the partners, the exit options, and the governance that keeps the arrangement reviewable. Anything broader is a national program, not an institutional one.

How should a large family office or institution structure AI exposure: direct deals, funds, or an internal capability?

Treat them as three different jobs. Funds buy diversified exposure and someone else's diligence, and you should still test that diligence. Direct deals require technical judgment you either have or must rent, plus concentration risk you should size deliberately. An internal capability is a different decision entirely, justified when AI changes how your own holdings operate rather than as an investment thesis. Most portfolios I see need clarity on which of the three they are actually doing before they need allocation numbers.

Is national transformation pressure a good reason to allocate capital, or are we chasing optics?

Alignment with a national program is a legitimate factor. It shapes demand, procurement, and partnerships for years, and ignoring it would be poor investing. It is not a substitute for a thesis. The test I would apply is whether the deal survives on its own economics if the policy environment stays exactly where it is today. If it only works with continued support, size it as a policy dependent position and say so in the paper that goes to the committee.

Where can we read your published work on the regional sovereign AI picture?

The Sovereign AI Landscape for the GCC and Lebanon 2026 is published at michaeljoseph.ai/sovereign-ai-landscape-2026/, maintained by a live monthly Sovereign AI Tracker. Both are public and sourced, so you can audit the reasoning rather than take a conclusion on trust. If you need the same analysis narrowed to a specific jurisdiction, sector, or investment question, that is a foresight engagement, and it starts with a working session on the decision you are trying to make.

Related: The 2026 institutional landscape (research) · Sovereign AI for the GCC · Strategic foresight · Selected work · All services