Pakistan's National AI Policy at one year

On 30 July 2025 Pakistan launched a National AI Policy with six pillars, a dedicated national fund, and a target of one million trained professionals by 2030. Twelve months on, the useful question is not whether the policy was ambitious. It plainly was. The useful question is what has been built, what has drifted, and what a Nordic reader learns from watching.

The proposition worth testing is this: can a large state with limited AI infrastructure compound literacy faster than a small state with abundant infrastructure. If the answer is yes, a good deal of received wisdom about human-capital sequencing in AI development is wrong.

The evidence

The policy sets out six pillars — Awareness and Readiness, Adoption, Innovation and Research, Skills and Training, Data and Compute, and Sustainability — alongside a dedicated National AI Fund and the one-million-professionals target for 2030 ^[1].

The starting conditions are demanding. Pakistan has a population of approximately 240 million and a labour force of approximately 80 million, with per-capita GDP roughly one-thirtieth of Norway's. It also has a large, young, English-competent technology workforce and a substantial diaspora, including in the Nordics.

The Norwegian comparator is the inverse on almost every axis. Norway has 5.5 million people, effectively universal digital-government infrastructure, and a public sector employing approximately 880,000 people ^[2]. It operates inside the OECD governance frameworks and inside the EU AI Act through the EEA.

What is actually comparable

More than the headline numbers suggest.

Both are single-language public administrations in practice — Norwegian in one case, Urdu with substantial regional languages in the other. Both have articulated national-scale AI literacy ambition at the level of the head of government rather than the ministry. Both are building against a binding legal floor or a close analogue of one. And neither is training a frontier model, nor plausibly will.

That last point is what makes the comparison honest. Strip away frontier capability and the question in both countries reduces to the same thing: how fast can a state make its own public administration competent at using systems it did not build.

The first-year read

In twelve months, the policy has produced a national narrative, a naming of pillars, and an aspirational target. Those are real outputs. National narratives are how large states coordinate, and a named pillar structure gives officials something to organise budgets around.

The gap between narrative and disbursement is where the next twelve months live. The leading indicators are not the aggregate pledged figure, which is the number most often quoted and the least informative. The indicators are the Fund's mandate document, its disbursement rules, and the criteria for the first cohort. A fund with a published mandate and named recipients is an instrument. A fund with a pledged total and no disbursement rules is a press release with a balance sheet attached.

This is not a Pakistan-specific failure mode. It is the standard failure mode of national AI strategies everywhere, including in Europe, where several member-state strategies published in 2019 and 2020 remain narratively intact and operationally thin.

Why the Nordic reader should care

Two reasons, one structural and one specific.

The structural reason is the analogue itself. If Norway and Pakistan can compound AI literacy on similar timescales once you adjust for scale, then human-capital AI investment behaves differently from infrastructure investment, and development finance is currently sequencing it wrong — funding compute and connectivity ahead of institutional competence on the assumption that competence follows hardware. If the analogue holds, competence is the constraint and hardware follows it.

The specific reason is that the bilateral is not hypothetical. Norway's Pakistani-heritage population is approximately 40,000 people. There is an existing bilateral relationship at government level, existing academic ties, and a diaspora with professional standing in both jurisdictions. The channel exists. Almost nothing is being sent through it.

What we would need to see to call it working

Three things, in order.

A published Fund mandate with disbursement rules. A named first cohort with published selection criteria. And a second-year evaluation conducted by someone other than the implementing body.

Any one of these on its own is weak evidence. All three together would put Pakistan ahead of most European member states on the accountability of national AI spending, which would itself be the story.

The bear case

If the Fund has not disbursed to a named first cohort by the second anniversary of the policy, 30 July 2027, then the policy remains an announcement rather than an instrument, and the analogue tells us nothing except that large states can write documents.

We will say that when it happens. We will also say it about the European strategies that fail the same test, which several will.

What ImpactLab is doing

The Norway–Pakistan publication track is the answer in publication mode. It is analysis, published annually, with a named diaspora and academic contributor list.

No institutional partnership counterpart has been named, and we are not claiming one. The relationship is one of published analysis, not a memorandum of understanding, and we will describe it that way until something is signed in public.