Waqar Uddin

Pakistan's National AI Policy 2025: A Pillar-by-Pillar Read

May 6, 2026 (3m ago)39 views

Pakistan now has a National AI Policy. It runs 28 pages, signed off by the Ministry of IT & Telecom, structured around six pillars with most deadlines in 2026 and 2027. Some of it is well thought through. Some of it follows the standard policy pattern of naming an objective and assigning it to a body that does not yet exist.

What follows is a pillar-by-pillar read for people who do not have time to chase the PDF.

Disclosure: I work at Jazz as Principal Evangelist Cloud & AI. This article is written as personal industry analysis. Views are my own.

Pillar 1: AI Innovation Ecosystem

The funding architecture. A National AI Fund (NAIF) draws at least 30 percent of Ignite's R&D Fund, on a perpetual basis under the Telecom Reorganization Act. Around it, a network of Centres of Excellence in AI.

Tying NAIF to a perpetual revenue source is the rare design choice that survives a change in government. Effectiveness will turn on whether the NAIF Committee actually convenes within two months of enactment, as mandated, and whether disbursement criteria become public. Without published criteria, the fund becomes another grant pipeline that everyone in the sector talks about and nobody can navigate.

Pillar 2: Awareness and Readiness

Big numbers, named owners.

The architecture leans on existing institutions rather than inventing new ones, which is the right call. The binding constraint is not headline numbers but trainer quality, content currency, and assessment. The 10,000 trainers target is the one to watch. If those trainers are themselves trained by CoE-AI to a serious standard, the rest of the pipeline is credible.

Pillar 3: Secure AI Ecosystem

The regulatory layer.

Governance bones are right, and a public register of state AI systems is rare in our region. Placing the AI Directorate under NCPDP signals a privacy-first lens. That is conservative but defensible. The cost is that AI policy decisions filter through a body whose primary remit is data protection, which can slow dual-use cases. The sandbox is the test. Sandboxes work when they move at startup speed. They fail when they replicate the approval-cycle dynamics of the institutions sponsoring them. Worth flagging: Pillar 5 and the Action Matrix later put the same sandbox target at 50 enterprises, not 20. The policy does not agree with itself on this number.

Pillar 4: Transformation and Evolution

Ten priority sectors for AI adoption.

Plus an AI maturity model in 50 institutions by 2026 and a Ranking Management System to evaluate AI solutions for public adoption.

The priority list is the right shortlist. Healthcare digitisation, agricultural advisory, and governance are exactly where AI compounding actually happens here. The text is thinner on funding. Sectoral roadmaps without ring-fenced budgets and named secretary-level sponsors will not survive a fiscal year.

Pillar 5: AI Infrastructure

The most ambitious pillar, and the most fragile.

Naming public cloud and open-source as default rails, rather than insisting on a sovereign-stack-from-scratch, is the correct posture for a country with our capital position. Distributing compute to 100 academic institutions also avoids concentrating national compute in one Islamabad facility nobody outside the federal capital can access. The Action Matrix later revises the same target up to 150 institutions, another instance of the policy's numbers moving depending on which section you read.

What is missing is the part that decides whether any of it gets built.

⚠️

No budget figure for the national compute grid. No mention of GPU server import duties (currently 48 percent). No power tariff line for HPC and data centre operations. No specification of terms for "industry-driven data centres." No anchor customer or co-investment vehicle named.

Compute is a procurement, energy, and tax problem more than a policy problem. The grid will succeed or fail on those line items, not on vision text.

Pillar 6: International Partnerships

Standard international cooperation language.

Useful but not where the policy differentiates. Most countries have something like this section. The actionable parts are FDI in AI ventures, capacity exchange, and standards alignment. The rest is scaffolding.

Governance: who actually does what

Single point of accountability is the Federal Minister IT. Worth knowing if you are tracking whether the policy is moving.

What to watch

Three execution signals will tell us whether this is a real policy or a shelf document: NAIF disbursement, the AI Directorate standing up under NCPDP, and infrastructure procurement getting real RFPs with real budgets attached.

I have since gone back and checked all three, against public reporting and against Pakistan's own approved federal budget documents, line by line. Short version: a Rs283 billion figure the government itself never actually approved, a funded AI programme almost nobody reported on, and a regulatory pillar that quietly changed its own foundation without saying so. Full scorecard, sourced to every claim, in the follow-up: Pakistan's AI Policy: What the Budget Books Actually Show.

Where this fits

Part of an ongoing series on Pakistan's cloud and AI infrastructure from a practitioner's perspective. Companion piece: Pakistan's AI Policy: What the Budget Books Actually Show. Previous posts: Picking Battles in the AI Stack, Pakistan's 5G Moment Is Also a Sovereign Cloud Moment, and Beyond Hyperscalers.