AICA head Norov: Pax Silica’s binary AI choice does not fit Central Asia
Pax Silica may be designed to secure trusted technology chains, but an exclusive US-or-China choice conflicts with Central Asia’s strategy of diversifying risk, says Vladimir Norov, a former Uzbek FM and former Shanghai Cooperation Organisation secretary-general, who now chairs the Central Asian Association for Artificial Intelligence.
A reported draft letter from the US State Department has turned an ostensibly non-binding technology partnership into a potential loyalty test. Reuters reported that Washington is preparing to warn the 35 signatories to June’s AI Opportunity Statement that participation in competing initiatives could exclude them from the US-led Pax Silica coalition.
The letter was an undated draft when Reuters reported it, yet its purported political message is clear: Washington increasingly views AI cooperation as a strategic trust test rather than a menu of projects partner governments can choose from.
For Central Asia, the signal reaches further. Kazakhstan, Kyrgyzstan, Tajikistan and Uzbekistan were among the 29 countries that signed an agreement establishing the World Artificial Intelligence Cooperation Organisation (WAICO), headquartered in Shanghai and orientated towards the Global South.
Read more: Kazakhstan faces US pressure over membership in China’s WAICO
Of the four, only Kazakhstan is officially in both camps, but any Central Asian capital seeking deeper technology ties with the US must now be mindful of the exclusivity clause, Vladimir Norov explained in a conversation with Kursiv Uzbekistan.
«It is realistic to expect Washington to become increasingly demanding about where its partners stand on certain sensitive technologies. It is much less realistic, however, to expect Central Asian countries to abandon their multi-vector foreign policies and choose one technology ecosystem at the expense of the other,» Norov stated.
Two offers, not two identical blocs
The Pax Silica Declaration covers frontier models, communications networks, compute, semiconductors, advanced manufacturing, logistics, mineral processing and energy. It calls for reducing excessive dependencies and protecting sensitive technology and critical infrastructure from undue access or control.
But treating Pax Silica and WAICO as interchangeable rival memberships obscures what each is selling. Norov put the difference in two short commands. The American approach is to secure the technology ecosystem, while the Chinese one is to expand access to it.
The US dominates the AI stack with advanced chip design, large capital pools, top labs, and unmatched data-centre capacity. China’s pitch to the Global South emphasises lower-cost technology, open-weight models, infrastructure, training, and participation in rule-making with plans for 5,000 AI training places for developing countries over five years, as well as cooperation centres linked to regional organisations, including the SCO.
Norov has warned against confusing accessibility with charity:
«This does not mean that China is acting purely out of altruism. By providing infrastructure, equipment, AI models, training and technology platforms, Beijing gains clear strategic advantages. The wider use of Chinese technologies also helps to spread Chinese standards, products and institutional practices.»
Neither ecosystem is geopolitically neutral, and the relevant question for Central Asia is which dependencies it can safely accept, at which layer of the technology stack and on what terms.
Cost is becoming an influential factor
An exclusivity demand might be more persuasive if one side had a clear and lasting technological lead. Stanford’s 2026 AI Index says leading US and Chinese models have traded places several times since early 2025. By March 2026, the top US model led its Chinese counterpart by only 2.7%. The US still produced more notable models in 2025 — 59 compared to China’s 35 — but model performance is no longer the only contest.
«The key question is therefore no longer simply who has the most powerful AI. It is increasingly about who can make advanced AI cheaper, accessible and widely available,» Norov argued. «That could determine not only technological leadership, but also geopolitical influence in the global AI race,» he said.
Multi-vector policy is risk management
In the meantime, no Central Asian state can currently supply every part of the stack on its own: electricity, critical minerals, cutting-edge chips, cloud infrastructure, models, research networks, capital and talent, Norov told.
«Central Asian countries do not have sufficient resources to build a complete modern AI ecosystem entirely on their own,» he admitted.
Read more: Central Asia risks becoming a consumer, not creator, of AI – World Bank
Central Asia’s reluctance to commit to one side is sometimes seen as an attempt to preserve its multi-vector approach. Yet this interpretation overlooks both the region’s geography and its economic interests.
Norov, himself a seasoned career diplomat, maintained that the foreign policy model of Kazakhstan, Uzbekistan and other Central Asian countries has traditionally been based on diversification rather than exclusive alignment with a single external partner.
«Their goal is not so much to balance ideologically between the United States and China as to avoid excessive dependence on any one external power while also attracting investment, technology, infrastructure and access to different markets,» he concluded.
