Why data secrecy is Pakistan’s biggest AI liability
T
he lesson from our experience is that in technology, timing is the true barrier to entry. We missed the global broadband and tech incubator wave of the 2000s because the 3G/ 4G spectrum auction was about a decade too late.
The government of Pakistan appears bullish on AI. It has upped the ante with a $1 billion investment pledge (over the next 4 years). A freshly minted National AI Policy is ready for rollout and though implementation is bound to go through several false starts as provincial governments warm up and a National Council takes shape, institutional momentum will likely catch up to policy priority. When that happens, we will run into a far more fundamental problem: you cannot build AI without data and Pakistan’s data is locked away.
The public sector is a highly data rich environment. This is true not just of the major data lords such as the NADRA which maintains one of the world’s largest biometric databases, covering most of Pakistan’s population. The FBR, the SBP and the ECP hold comprehensive, granular data on how citizens spend, save and vote. The PBS and the BISP know all that is worth knowing about consumption patterns, living standards, employment rates and demographic shifts. However, the raw micro-data sits behind selective paywalls.
Then there’s the quantum realm of specialised studies, surveys, portals and dashboards commissioned by governments through IFIs, UN agencies and foreign contractors that never see the light of day. Results are presented in darkened conference rooms and then relegated to the hard disks of forgotten desktops. The data– digitised, organised and shareable – exists. The will to share it, does not. This makes data the most important prisoner in Pakistan today.
Does this culture of secrecy have an upside? Does it ensure accuracy? No. Without public scrutiny, data compounds its errors. The $11 billion discrepancy in import data from 2023-25 caught by the IMF turned into an international embarrassment. Does it ensure security? Not really. From 2019-2023, over 2.7 millioin Pakistanis were reportedly robbed of their personal data from NADRA. It was eventually sold to buyers in Romania and Argentina.
The downside, of course, is the incalculable opportunity cost i.e. the potential of open access for researchers, entrepreneurs and innovators. How many scientific breakthroughs were kept from us? How many policy problems could already stand solved? How many tech start-ups will we never see? We have starved human intelligence of needful data and it looks like we’re about to do the same to artificial intelligence.
Things are changing. Provincial land records are now digitised and searchable on online portals. The Planning Commission has just published past and on-going PSDP project datasets online. Karandaaz has just launched a data aggregator, platforming datasets on finance, commodity trade and other economic stats. The National AI Policy recognises the need to liberate data from existing systems such as “WAPDA, the FBR, health sector, the NADRA, education and ministries etc” to nourish AI.
The problem is that the degree and pace of this change lags behind the world. It’s not a coincidence that Pakistan ranks low on global indices for both data openness (150/198) and AI readiness (81/195) for 2025. In India, AIKosh has already on-boarded nearly 6,000 datasets and 250 AI models to populate its national repository. Singapore’s data.gov.sg portal offers over 4,000 datasets from 69+ government agencies completely free for users and businesses. The European Cancer Image Data Space hosts 60 million (and counting) anonymised cancer images and annotations.
The lesson from our experience is that in technology, timing is the real barrier to entry. We missed the global broadband and tech incubator wave of the 2000s because the 3G/ 4G spectrum auction was about a decade too late. We caged our e-commerce in a cash-on-delivery ecosystem while the fin-tech revolution passed us by in the 2010s. In each case, early movers moved and won – we didn’t and lost.
Pakistan cannot now be a latecomer to AI. The stakes are simply too high. We either ride the AI wave past the middle-income trap or be swept away to further economic stagnation, inequality and low competitiveness.
The writer is a governance and public policy professional with over 20 years of experience in the development sector. He has worked with donor agencies such as the USAID and the FCDO.
Leave a Reply