ThinkPatternGet the app
Perspective
TECHNOLOGY · JUL 19, 2026

Between Two AI Walls, the Global South Looks for a Door

China's WAICO offers governance without technology. America's Pax Silica offers technology with conditions. Neither gives developing nations what they actually want.

On July 1, 2026, China's new AI export controls took effect — 34 articles signed by Premier Li Qiang that restrict model-weight downloads, ban unauthorized cross-border talent transfers, limit API access, and designate AI model leaks as national security offenses [1][2]. Fifteen days later, Xi Jinping addressed the founding conference of the World AI Cooperation Organization in Shanghai and offered a diagnosis of what was wrong with global AI governance.

We should jointly oppose overstretching the national security concept in the field of AI and placing one country's security over that of others. — Xi Jinping

The dates do the work. No outside accusation of hypocrisy is required; the calendar is the argument. China launched WAICO on July 16 with 29 founding members, UN Secretary-General António Guterres in attendance, and a promise of 5,000 AI training slots for developing countries over five years [3][4]. It launched it with a critique of technological gatekeeping that its own export controls, effective 15 days earlier, had just codified into law. The contradiction is not a slip — it is the architecture. WAICO offers governance inclusion to everyone and frontier technology to no one. The US has its own wall, and it is built differently. Pax Silica, launched in June with 35 nations and over $250 billion in committed deals, is an explicit supply-chain exclusion mechanism: it diversifies semiconductors, critical minerals, and compute away from China and channels them to partners the US deems trustworthy [5]. Deputy Secretary Landai made the logic plain.

Pax Silica exists to keep these technologies and the future growth of all of us in trusted hands. — Christopher Landau

The US stack offers real money and real technology — but only to countries inside the perimeter, and only as long as Washington keeps the gate open. India learned what that means when Anthropic, acting on a US directive, suspended advanced model access to Indian users. India's MEITY Secretary responded bluntly.

We sought an understanding of how exactly the US is looking at this particular aspect and what their concerns are and how, in the future, this could be a reliable source of technology, because if it is something which is to be used and made available, we can't have abrupt cutoffs. — Maya Sundarakrishnan

China's wall is subtler. WAICO is not a technology-transfer body; it is a governance body. It offers rules, training, and a seat at the table — not model weights, not API access, not the frontier models China's July 1 controls locked down. Guterres lent the enterprise UN legitimacy, arguing that AI governance cannot be left to a handful of countries or companies [3]. But WAICO is headquartered in Shanghai, not Geneva, and its preparatory documents route cooperation through the Shanghai Cooperation Organization and BRICS — embedding the body in existing China-led infrastructure, not the UN system [6]. China secured the UN Secretary-General's presence and endorsement for WAICO, lending it multilateral legitimacy the US stack never sought, while routing its cooperation through institutions Beijing already leads. The US, for its part, had already made its position on UN-led AI governance explicit: at the New Delhi AI Summit in February, the administration flatly rejected any UN role in governing the technology [7]. So the two poles are now symmetric in structure and opposite in offer. Both wall off their frontier technology. Both use national security as the justification. The US gives technology to trusted partners and withholds it from adversaries — but the trust is revocable, as India discovered. China withholds frontier technology from everyone and offers governance inclusion to everyone — a seat at the rules table, not a key to the model lab. At the UN Global Dialogue on AI Governance, China's representative framed the choice in language that could have been drafted for a non-aligned movement summit: digital sovereignty as the right to choose AI products without being coerced into taking sides [8]. India is where the divide is visible from inside, not from either capital. It joined Pax Silica. It signed $210 billion in deals with Reliance and Adani. It got cut off. And at that same UN dialogue, India's minister used language nearly identical to China's, calling the forum a moment where nations must choose consensus over conflict [8]. India is not defecting from one camp to the other. It is hedging — inside both stacks and comfortable in neither. And it is not alone. Canada and Germany launched a Sovereign Technology Alliance in February explicitly to reduce strategic dependencies on both US and Chinese platforms [9]. India and Germany signed a separate AI pact the same month. Netherlands Prime Minister Schoof captured the logic: the US and China are competing so intensely with each other that other countries must come together and develop their own approach [10]. These are not abstentions from the great-power AI contest. They are separate architectures — a third stack, still under construction, built on the premise that neither Washington nor Beijing offers what developing and middle-power nations actually want: technology access without political conditions. What makes this more than a diplomatic taxonomy is the technical substrate beneath it. The Stanford AI Index confirmed in April that the US-China AI performance gap has effectively closed, with China producing 74.2% of global AI patents and an 80% drop in AI researchers moving to the US [11]. The US-China Economic and Security Review Commission warned in March that 80% of US AI startups now use Chinese open-source base models, and Alibaba's Qwen has surpassed Meta's Llama in cumulative downloads [12]. China is building WAICO's institutional layer on top of a technical substrate that is already globally embedded. The governance body's reach is potentially self-reinforcing: the more the world builds on Chinese base models, the more natural Chinese-led governance becomes. But the open-source substrate also cuts the other way. If the models are already out there — downloadable, forkable, running on servers no government controls — then the institutional walls both superpowers are building may matter less than either capital believes. WAICO offers a seat at the rules table. Pax Silica offers a key to the model lab. What neither offers is what the Global South is already doing: taking the open models, building sovereign capacity on top of them, and looking for the door.


Sources
  1. 1. China Imposes Strict AI Export and Investment Controls
  2. 2. US and China Escalate AI Conflict Over Security and Exports
  3. 3. China Launches World AI Cooperation Organization in Shanghai
  4. 4. China Hosts World AI Conference and Pledges Global Training
  5. 5. US and 34 Nations Launch Pax Silica AI Initiative
  6. 6. China Accelerates Establishment of World AI Cooperation Organization
  7. 7. US Rejects UN Global AI Governance at New Delhi Summit
  8. 8. India and China Call for Inclusive Global AI Governance
  9. 9. Canada and Germany Launch Sovereign Technology Alliance for AI
  10. 10. India and Germany Sign AI Pact to Counter US-China Dominance
  11. 11. Stanford Report Says US-China AI Performance Gap Has Closed
  12. 12. US Commission Warns China's Open-Source AI Threatens US Leadership

Keep reading in the app

The full perspective, free in the app.

Download on the App StoreComing soonGoogle Play