The Pakistani government recently unveiled an AI-based system to digitally track and monitor the implementation of prime ministerial directives. The initiative is the latest in a series of measures aimed at positioning Pakistan at the forefront of the global AI wave. Other initiatives include the National AI Policy, the establishment of the Pakistan Digital Authority (PDA), a proposed $1 billion investment fund to support AI start-ups, the rollout of domestic AI data centres, the deployment of AI to strengthen tax collection, and efforts to use factory-generated data to train industrial robots.
Yet despite these advances, a major weakness remains at the heart of Pakistan’s AI drive: its AI stack is heavily dependent on foreign technology. The stack—which includes chips, servers, storage, cloud infrastructure, operating systems, AI models and applications—is built largely on proprietary components supplied by foreign, predominantly American, companies. This dependence leaves Pakistan’s AI ecosystem exposed to external commercial, technological and geopolitical pressure.
The pursuit of sovereign AI stacks has recently acquired fresh urgency. American efforts to restrict access to frontier AI models and advanced GPUs, its campaign to exclude Chinese AI products from global markets, and the Iran war have transformed AI from a source of commercial advantage into a strategic national-security asset.
Countries including France, India, Saudi Arabia, Japan and South Korea are moving towards sovereign AI stacks despite the enormous cost and time involved. Since none can achieve complete technological independence immediately, they are adopting interim strategies—using imported chips, foreign partnerships, open-source models and nationally controlled data centres—while steadily building domestic capabilities across infrastructure, cloud, models and applications. Pakistan is not immune from this global shift and must decide whether to manage its dependence strategically or remain permanently exposed to technologies controlled by foreign powers.
What makes this more pressing for Pakistan is that AI, unlike previous general-purpose technologies, has the potential to become the “mother” of all such technologies, reshaping productivity, industry, defence and state capacity. This means Pakistan must be prepared to commit adequate resources and pursue a deliberate long-term path towards sovereign AI or risk being left behind and compromised.
Pakistan desperately needs to move in this direction to guard against plausible scenarios that could threaten its national security. Recent conflicts have shown how foreign-controlled digital technologies can become instruments of warfare. Israel’s Lavender system was used to generate
targets in Gaza, while the mobile phones carried by the bodyguards and drivers of senior Iranian officials were reportedly hacked and tracked, exposing the locations of leaders who were subsequently targeted and killed. Consider another plausible scenario: if relations between Pakistan and the United States were to deteriorate sharply, what would prevent Washington from weaponizing its control over critical technologies—including Microsoft Copilot, cloud platforms and operating systems—to disrupt economic activity, restrict access to essential services or compromise sensitive state and commercial data?
Given AI’s pervasive, multi-purpose applications across the economy, government, defence and critical infrastructure, it must be treated as an integral component of Pakistan’s national security. Critics contend that building a complete domestic AI stack would be prohibitively expensive, take many years and force Pakistan to reproduce technologies in which it already trails global leaders.
That economic objection is not decisive because it treats AI as an ordinary commercial investment rather than a strategic capability. Pakistan should respond in two stages. First, it should adopt a low-cost interim strategy of layered sovereignty, using open-source models, affordable chips, domestically controlled data centres, diversified suppliers and secure national platforms to reduce immediate exposure.
Second, it should mobilise a long-term national effort comparable to its nuclear programme. Nuclear capability was also prohibitively expensive, technologically difficult and beyond Pakistan’s existing capabilities, yet it was pursued because national security outweighed short-term economic cost. Within roughly 25 years, Pakistan established a credible deterrent and achieved strategic parity with India.
The relevant question, therefore, is not whether sovereign AI is expensive, but whether the long-term cost of technological dependence is greater. Pakistan does not need more headline AI successes that deepen foreign dependence. It needs an immediate strategy of layered sovereignty, followed by a sustained national effort to build its own sovereign AI stack, even if that ambition takes decades to realize. The Government of Pakistan, through the Pakistan Digital Authority, should place this objective at the centre of its national AI agenda.
