Pakistan has joined China’s AI body. Now comes the hard part

Deputy Prime Minister Ishaq Dar spent last week in a Shanghai signing room, putting Pakistan’s name to the founding charter of the World Artificial Intelligence Cooperation Organization (WAICO), a new China-led body for AI governance. Twenty-eight other states signed alongside him, from Russia to Indonesia to Kazakhstan. No tanks moved. No treaty was renegotiated. But beneath the export controls and the compute gap separating American and Chinese frontier models sits a second, quieter contest: whose rules, whose standards bodies, and whose training pipelines middle-income states plug into as they build AI capacity. That contest just reached Islamabad, and it will reach dozens of other capitals in the years ahead. 

The case for WAICO membership is diversification, not a vote for Chinese AI over American AI. Pakistan already depends on China for infrastructure, telecom, and energy through CPEC; AI cooperation extends a relationship that predates the current technology race by decades. Islamabad gains knowledge transfer, joint research, and training pipelines it cannot easily obtain elsewhere, while maintaining that it remains free to buy American cloud services and adopt European regulatory templates. 

The argument is coherent. It is also familiar. Pakistan made comparable claims about non-alignment at Bandung in 1955, about its nuclear program in the 1970s, and about IMF programs run in parallel with Chinese loans. The claim to sit outside both camps has never been false. It has also never been free. Each past hedge produced roughly the same result: partial technology transfer, partial capacity built at home, and a widening gap between how sophisticated the partnership looked on paper and how competently it was used in practice. 

The WAICO agenda makes that pattern legible, though it also contains a genuine opportunity that the older hedges did not: education and human capital; applied AI in agriculture, health, and public administration; and language models for Urdu and regional languages. Of the three, the last is real and underserved: Pakistan needs Urdu-language models more than it needs another summit invitation, and no American lab is racing to build them. But the other two require institutions Pakistan has struggled to build regardless of which partner supplies the technology. A joint research program is only as good as the universities absorbing it. Applied AI in public administration is only as useful as the state capacity deploying it, and Pakistan’s civil service has resisted digitizing far simpler systems, from tax collection to land records, for a generation. 

The real risk is one Islamabad rarely states out loud: success depends on Pakistan’s own competencies and regulatory frameworks, not the generosity of the partnership. A government that could not build a functioning tax-filing portal without repeated World Bank interventions now holds a seat on the AI governance body. The gap between drafting a framework and running one has swallowed prior Pakistani technology initiatives. The Digital Pakistan Policy of 2018 was followed by a draft National AI Policy in 2023 that took two more years to be formally approved by the cabinet, in mid-2025. The government launched a sovereign AI cloud and a startup fund under it that October. Nine months on, there is little public trace of either operating. The pattern is not malice. It is the same institutional bottleneck each time, wearing a new label. 

The Pakistani state signs frameworks reliably and runs them rarely. The question worth asking is not who wins Pakistan’s signature. It is who answers the phone when the tool fails to ship.

-Javed Hassan

Pakistan’s rare tech successes (NADRA’s identity system, the Raast payments rail) prove the point rather than undercut it: each had one institutional owner. A joint AI framework, split across ministries and a foreign partner, has none. 

Widen the lens to the Gulf, and the contrast sharpens. Saudi Arabia and the UAE buy American chips through licensed Nvidia and Microsoft deals while building parallel partnerships with Chinese firms on infrastructure and applications. They can do this because they have sovereign wealth large enough to be courted by both Washington and Beijing on commercial terms, and technocratic institutions, the Public Investment Fund, Abu Dhabi’s ADQ, strong enough to absorb what they buy. 

Pakistan has neither the capital nor the institutions. It has geography, population, and a relationship with China that predates AI. Its hedge is closer to Bandung than to Riyadh: a rhetorical claim to autonomy sitting on asymmetric dependence. 

Washington tends to file Gulf AI diplomacy and Pakistani AI diplomacy under the same story: another front in a binary contest. They are not the same story. Gulf states buy optionality with capital they possess. Pakistan is trying to buy it with a relationship that, on its own, cannot supply what capital does. Chip export licensing carries real leverage in Riyadh, because Riyadh has an alternative worth choosing. It carries far less in Islamabad, where the alternative to Chinese cooperation is not American cooperation but no cooperation. 

The test of WAICO membership, then, is not which flag flew over the signing photo. It is whether Pakistan produces a working Urdu-language model, a usable agricultural AI tool, and a regulator that enforces data rules regardless of who built the system. On the evidence of the last four decades, the Pakistani state signs frameworks reliably and runs them rarely. The question worth asking is not who wins Pakistan’s signature. It is who answers the phone when the tool fails to ship. 

Javed Hassan has worked in senior executive positions in the for-profit and non-profit sectors in Pakistan and internationally. He is a former Senior Visiting Fellow at Fudan University, Shanghai.