Atif Ahmad LinkedIn ↗
← All signals

AI Governance June 29, 2026

Governing AI at the level of a country — where the money flows, and what you're really trading off

Most conversations about “AI governance” happen inside a single organisation. The harder, more consequential version is one level up: how does a country govern a general-purpose technology that touches health, defence, education, finance and public services all at once? That isn’t a policy. It’s a system — and like any system, it succeeds or fails on how well its layers connect.

Governance has to trickle down — and that’s where it breaks

National strategies tend to lead with ambition. The UK’s, for one, was framed explicitly around “innovation and opportunity” (review of the UK national AI strategy, 2021). Ambition is fine. The failure point is the handoff: principles set at the centre must become operational rules inside sector regulators, then real practices inside institutions and firms. When that trickle-down is weak, you get one of two pathologies — box-ticking that satisfies the letter and misses the point, or a vacuum where “the national framework” is invoked but nobody local can actually apply it.

And the binding constraint on that handoff is rarely the policy. It’s capacity. Governing AI in the public sector is, at bottom, a capability problem — the skills, processes and institutional muscle to actually assess and oversee these systems (AI Governance Capacity in Public Administration, 2024; AI Governance Strategies for the Public Sector, 2025). You cannot delegate governance to layers that lack the ability to exercise it.

Where central money should flow

Markets fund what pays quickly. A government’s job is to fund the connective tissue markets won’t — the things that only work at national scale and that make everything downstream possible:

  • Shared compute and public data infrastructure — so capability isn’t hoarded by whoever can afford it.
  • Foundational and public-interest research — the long-horizon work with no near-term buyer.
  • The talent and education pipeline — governance capacity has to be built, not assumed.
  • A national assurance and standards capability — the audit-and-certification backbone that lets the trickle-down actually flow. Much of near-term governance will run on standards and certification rather than hard law anyway (Kshetri, 2024).

Spread thin, this money achieves nothing — which forces the real question.

Where prioritisation has to happen

You cannot govern, or fund, everything at once. Prioritise on three axes: risk (high-consequence domains like health and critical infrastructure first), sovereignty (the capabilities a country genuinely needs to control for itself), and public value (where the upside for citizens is largest). Everything else can wait, or be left to the market.

The trade-offs nobody wants to name

Here’s the part strategies gloss over. These are genuine give-and-takes, not problems to be “solved”:

  • Speed vs. assurance — every month spent building trust is a month a competitor ships.
  • Central standards vs. local fit — uniformity is governable but blunt; sector-specific rules fit better but fragment.
  • Sovereignty vs. openness — national capability can harden into isolation, and global governance can quietly entrench whoever wrote the rules first — a risk the Global South is right to name (Decolonizing global AI governance, 2024).
  • Infrastructure vs. people — compute is visible and fundable; the human capacity to govern it is neither, and gets starved first.

A national AI strategy that won’t name its trade-offs isn’t a strategy — it’s a press release.

For three decades I built the systems that run national carriers, and the lesson transfers cleanly: you cannot govern what you cannot see, and the highest-leverage central investment is almost always the connective tissue — the standards, the assurance, the shared infrastructure — because that’s what turns a slogan at the top into a practice at the bottom. Govern the handoffs, fund the connective tissue, and be honest about what you’re trading away. The countries that do that will compound; the ones chasing the announcement won’t.

References

← All signals