Moinuddin Ahmad 

The recent passage of Pakistan's National Artificial Intelligence Policy is more than just a domestic industrial strategy; it also serves as a potential tool for foreign policy. The choices Islamabad makes concerning data localisation, the personnel managing the computing infrastructure, and the partners granted special access to the market will shape the country's strategic independence for decades to come. This policy requires a geopolitical framework rather than merely a technical plan. Major world powers and leading cloud providers are already viewing data, computing power, and AI talent as means of influence. In Pakistan's case, current policy decisions could decide whether the nation becomes a rule-taker or a rule-maker. Ultimately, the stakes are not solely economic and technological but also existential, touching on national sovereignty and diplomatic influence.

The signing of the National AI Policy in 2025 by the cabinet creates financial resources, infrastructure priorities, and an explicit mandate to establish partnerships with other countries, indicating that Pakistan would ramp up quickly in scale, compute, skills, and research. Meanwhile, the government has taken the step of monetising excess power for the digital industry, with announcements of up to 2,000 megawatts for bitcoin mining and AI data centres, which transform physical energy capacity into bargaining capital for foreign investors. The technology question is no longer independent of geopolitics due to these two policy streams of AI strategy and energy allocation.

Three short-term strategic risks. To begin with, legal uncertainty regarding cross-border data transfers and the Personal Data Protection regime poses a threat to the emergence of foreign data enclaves. Pakistan. The drafts of the Pakistani data-protection framework suggest localisation and registration but lack enforcement mechanisms and adequacy processes, which is the ideal place to speculate on contractual terms of exploitation by strong vendors.

Second, highly complex global supply chains and cloud markets can lead to vendor lock-in. Specialised AI chips, data centre design, and managed AI stacks are hyperscale cloud providers and a few producer states. Commercial contract terms and technical dependencies (firmware provenance, locked APIs) may convert economic relationships (when a narrow vendor base) into a strategic advantage for Pakistan if the compute and procurement model favours a small group of vendors. Third, there are domestic security and rights risks due to governance gaps: the rapid adoption of AI based on independent regulation, access history, and judicial controls increases the potential for mission creep in surveillance technologies. The weak privacy protections and the need for a stronger, more serious data-protection force have already been noted in the civil society reviews of the AI Policy.

Five useful pillars of a geopolitical design. The asymmetric dependence can be mitigated through available policy tools that balance openness and sovereignty in Pakistan.

1.    Tiered data sovereignty. Categorise datasets by sensitivity, require onshore processing of sensitive datasets (citizen registries, biometrics, health data, defence-contiguous telemetry), and allow auditable exports of non-sensitive commercial datasets under adequacy agreements. Such a hybrid solution maintains research team working relationships and cloud economies without loss of control over strategic assets.

2.  Conditioned compute as well as energy allocation. Condition special energy with quotas (e.g., 2,000MW programme) on open government governance standards: audited uptime guarantees, a time-limited local employment requirement, verifiable technology transfer, and escrowed administrative control over vital services. Assign capacity to competitive, geopolitically sensitive auctions rather than to informal, first-mover transactions.

3.    Energetic procurement and multi-vendor architecture. Multi-vendor procurement should also be mandated, and open-source reference stacks for government workloads, as well as national mirror nodes for sensitive datasets, should be required. Establish a tiny yet professional Compute Trust office to certify the supply chains of hardware to ensure that they have been produced by a certified supplier, confirming the provenance of their firmware, and running adversarial resiliency tests - cheap to set up now, but expensive to buy back later.

4.  Capability-linked partnerships. Tie all things to co-development: audited codebases, shared IP policies, investor-funded upskilling goals, and open legal discussions on external access to data should be preceded by any foreign partner before gaining access to any data. Issue set tranches from the National AI Fund based on conditional milestones, in a way that encourages the transfer of capabilities rather than just deployment.

5. Artificial intelligence diplomacy and technical representatives. Include technical expertise within the diplomatic post to agree on the data sufficiency, cross-border research reciprocity, and export-control standards. Demand multilateral systems of dispute resolution to avoid cross-border access orders that are unilateral and to avoid extraterritorial claims to Pakistani data.

Evidence of comparators- what works. Pakistan does not have to work it out again. The adequacy dialogue in the EU demonstrates the potential of limited, legally binding equivalence to facilitate cross-border flows and to afford citizens. The OECD best practices for procurement and escrow requirements in some high-tech tenders can illustrate how to prevent lock-in through contract design. Another factor that should not be overlooked is the domestic precedent: countries that established separate data authorities and supply-chain test laboratories early found it easier to maintain policy space as geopolitical pressure increased.

Response to the critics: being open with strings. Others will claim that circumstances retard investment. The bilaterality of such a counter-argument is asymmetric dependency: fast, unconditional openness can be a draw for capital, but can also lose long-term bargaining power. A conditional-openness model - providing access to market in exchange for verifiable transfer, audit, and co-development can maintain pace and sovereignty.

Short-term measures that can be taken. Publicise a cross-ministerial list of sensitive datasets and custodians; turn any special energy or compute assignments into time-limited, audited contracts with local capability conditions; legalise and resource a separate information-defence agency with expressive enforcement and audit competence; and provide diplomatic missions with technical negotiators to negotiate equitable research and cloud arrangements.

The policy of AI in Pakistan will not be evaluated by the ratio of data centres to be hosted, but by the data owner: the information, the processing, and the policies. The discussion about AI as a geopolitical tool alters the discourse from passive to active governance. This is pressing: the time frame during which Pakistan can formulate contractual norms, source competitively, and develop internal competencies is not in years; it is measured in months. Parliament should review and require the tiered-sensitivity approach in a timely manner; the government should turn ad hoc energy distributions into competitive, conditional tenders; and diplomatic posts must promptly begin negotiating adequacy and co-development deals with key partners. As a conditionally open nation, procurement protection and combined AI diplomacy enable Pakistan to achieve rapid digitalisation while preserving its sovereign policy space. Now it is time to create AI as statecraft.