Pakistan's AI Policy: What the Budget Books Actually Show

@kawishwaqar|July 18, 2026 (3w ago)12 views

I wrote a pillar-by-pillar read of Pakistan's National AI Policy in May and set myself a checkpoint: three execution signals to watch, and if two of three moved by Q3 2026, the policy was real rather than a shelf document. That checkpoint has arrived. I checked all three against public reporting and, this time, against Pakistan's own approved federal budget: the Annual Plan 2026-27, its Highlights, and the final PSDP 2026-27 book.

The press-reported numbers around this policy do not match what the government actually approved to spend.

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

The headline number doesn't hold up

Pakistani press, APP's wire service among them, reported a Rs283 billion, five-year "National AI Ecosystem Development Programme" running 2026 to 2031. I cited that figure myself in the original piece. It is not what the Central Development Working Party actually approved.

The PSDP 2026-27 book, IT & Telecom Division, lists it as a New Scheme, G.Sl. No. 488, "National AI Ecosystem Development Program," approved by CDWP on 18 June 2026 at a total cost of Rs13,000 million. Of that, Rs12,862.191 million is foreign-aid financed, meaning the domestically-funded portion of this "Rs283 billion" flagship programme is under Rs140 million. The FY2026-27 allocation is Rs185.064 million, split Rs100 million foreign and Rs85.064 million local, which matches the Rs185 million figure reported in the press almost exactly. That specific number was right. The Rs283 billion frame around it was not.

I don't know whether that gap is a genuine drafting inconsistency, a vision document that got approved in a much smaller first tranche, or a number that got repeated in press releases faster than it got reconciled with what CDWP actually signed off on. What I know is that the number CDWP approved, in the government's own current-year budget book, is roughly 4.6 percent of the number that ran in the papers.

A second programme nobody reported on

The same PSDP book has a separate, on-going scheme, G.Sl. No. 483: the National AI Advancement Initiative (NAAI), approved by the Departmental Development Working Party on 30 June 2025, total cost Rs534.639 million, with Rs524.639 million allocated for FY2026-27. This is a real, currently funded programme, already a year into its approval, and I could not find a single press article naming it. It does not appear anywhere in my original search of Pakistani or international coverage. It surfaced only because the government's own official acronym list, in the Annual Plan's abbreviations annex, spells out both NAAI and NAEDP as distinct entities.

NAIF itself is not in this budget at all

The National AI Fund, the mechanism the original policy leans on hardest, does not appear in the PSDP or the Annual Plan by name. That is not an oversight on the government's part. NAIF draws its money from a permanent 30 percent share of Ignite's R&D Fund, itself financed by a levy on telecom operators under the Telecom Reorganization Act, not from the federal development budget. It sits structurally outside the PSDP process entirely. NAIF, NAAI, and the AI Ecosystem programme are three separate funding vehicles, not one pot counted three ways, and no public disbursement record exists for any of them.

The rest of the scorecard

By my own May 2026 test, that is zero of three signals clearly moved, and infrastructure procurement, the third, has not moved at all.

What's genuinely real

Indus AI Week ran February 9 to 15, with the Prime Minister attending, a USD 1 billion AI commitment announced through 2030, and the Islamabad AI Declaration issued, a document not in the original policy at all. AI Seekho 2026, a MoITT, Google, Telenor, and Innovista training initiative, launched against the Pillar 2 workforce targets. Pakistan became a founding member of the World AI Cooperation Organisation in July. None of that is nothing. All of it is visibility rather than the institutional and financial plumbing the policy actually specified, and Dawn's own reporting found the same pattern in February: of the six pillars, only Awareness and Readiness had visible movement, the other five negligible, and on AI Infrastructure specifically, "currently no authority responsible for regulating or monitoring" it.

The candid take

I do not think this is a policy that failed. I think it is a policy that got a good document, a real if confusingly-named set of funding lines, and a weak follow-through mechanism, and the three are being allowed to drift apart in public view while the ceremonial calendar, summits, declarations, international memberships, fills the gap where an implementation record should be.

The NCPDP situation is the clearest tell on governance. The policy's entire regulatory pillar hangs off a commission that requires an Act of Parliament nobody has passed, and rather than force that fight, the government appears to be quietly building the same authority into the Pakistan Digital Authority through a different policy instrument. That may well be the more pragmatic path. Running it as an unannounced pivot instead of an acknowledged change with a revised timeline is what erodes trust.

The Rs283 billion figure is the equivalent tell on money. A number that large, repeated across multiple outlets, sitting twenty times higher than what CDWP actually signed off on, is not a rounding difference. It is either a communications failure or a budget that has not caught up with its own press coverage. Either way, it is now on the public record, in the government's own budget book, at Rs13 billion.

A constructive action plan, in the order I would actually push these:

1. Publish a single reconciliation note explaining the Rs283bn-to-Rs13bn gap. Whether it is phased approval, a drafting error, or a downsized scope, say which. Silence on a 20x discrepancy costs more credibility than the discrepancy itself.
2. Notify the AI Council now, publicly, with its full member list. It has been in nomination limbo since at least February. Every other execution signal in this policy nominally answers to a body that does not yet formally exist.
3. Reconcile the NCPDP/PDA question in writing. Either fast-track the NCPDP Act, or formally amend the AI Policy to name the Pakistan Digital Authority as the AI Directorate's home once the Data Governance Policy clears cabinet.
4. Publish a quarterly disbursement dashboard for NAIF, NAAI, and the Ecosystem programme, with named grantees and amounts. Budget lines without a visible trail of who received money are indistinguishable from budget lines that never got spent.
5. Fix the sandbox and institution-count contradictions before the first cohort launches. Twenty versus fifty enterprises, one hundred versus one hundred fifty institutions: these stop being typos the moment a real applicant gets rejected by citing the "wrong" number.

None of this requires new legislation beyond the NCPDP question, and none requires new money beyond what has already been announced. It requires someone to be accountable for turning press releases and budget books into a record that agrees with itself.

Where this fits

Part of an ongoing series on Pakistan's cloud and AI infrastructure from a practitioner's perspective. Companion piece: Pakistan's National AI Policy 2025: A Pillar-by-Pillar Read. Previous posts: Picking Battles in the AI Stack, Pakistan's 5G Moment Is Also a Sovereign Cloud Moment, and PISF Closes the Hosting Layer in Pakistan's Sovereignty Stack.

• Website: https://waqaruddin.com
• Medium: https://medium.waqaruddin.com/