On July 16, 2026, at the World Artificial Intelligence Conference in Shanghai, representatives from 29 countries in the Global South signed an agreement establishing the World Artificial Intelligence Cooperation Organization (WAICO), an intergovernmental organization proposed by Chinese President Xi Jinping and headquartered in Shanghai.
To assess the potential impact of the new organization, we approached several experts to reflect, among other things, on two key questions:
-
What does WAICO’s establishment mean for the future of global AI governance?
-
Whose and what interests does the organization serve?
Michael Caster, Head of Global China Program at ARTICLE 19
In his opening remarks at the conference in Shanghai on July 17, Xi reiterated the CCP’s prioritization of so-called “true multilateralism,” language China increasingly uses to justify establishing parallel institutions under Beijing’s leadership. China’s multilateralism equally rests on cyber sovereignty as a norm, which entails a model of global digital governance that is functionally unencumbered by international human rights law and internet freedom principles. WAICO is emblematic of China’s ambitions to embed its own institutions into the global AI and emerging technology governance ecosystem.
Institutions matter not only for the principles they proclaim or succeed at influencing directly, but also for their ability to convene governments, train officials, set technical standards, coordinate research, and build durable networks. These activities gradually shape what becomes accepted as international best practice.
In that sense, while certainly among its ambitions, China does not need WAICO itself to fully supplant existing fora any more than for it to become a significant venue for the normalization of China’s approaches to AI governance, which participating members can then actively promote in other fora that remain more influential. China is not just building new institutions; it is also seeking to selectively coopt the more state-centric strands of the existing system over multistakeholder ones. For example, China continues to tout the importance of engagement through UN mechanisms for digital governance but prefers the more multilateral elements of the Global Digital Compact over the more multistakeholder WSIS+20 process [the 20-year review of the World Summit on the Information Society] and the Internet Governance Forum.
The same week WAICO launched, China reinforced such global posturing with its Position Paper on Global Cyber Governance in the Digital Intelligent Age, presented to the UN’s Global Mechanism on ICT Security. Again, China promotes respect for cyber sovereignty, allowing states to independently choose their AI technologies, and upholding multilateralism. In practice, however, this means deference to state control of information and the legitimization of authoritarian tech stacks as alternative models to rights-based frameworks.
Europe should take this seriously. First, as the EU works to embed rights-based principles, from its own AI Act to broader contribution in global AI governance, China’s parallel-institution strategy competes directly for the same diplomatic and normative influence. Second, this is not a hypothetical contest at Europe’s periphery. Serbia is among the 29 global founding members, along with more usual suspects like Belarus and Russia. A European state lending early legitimacy to an organization built to normalize Beijing’s alternative to rights-based AI governance should be a wake-up call.
For now, China’s bid for AI governance leadership remains more aspirational than impactful, but the cost of dismissing it is high, and Europe should especially be watching closely which states follow Serbia in membership or engagement.
Veronika Blablová, Research Fellow at CEIAS
WAICO seems to be mostly read as a norm-setting body, which might be the wrong starting point, as governance and standards cooperation is only one of its pillars. Chinese rhetoric emphasizes development, innovation cooperation, and bridging the AI divide, while also building the principles of AI governance, standards, and norms. Although little is known about the platform’s functioning, funding, and more specific initiatives, Xi announced that China will provide AI training and seminars to developing countries, create cooperation centers with other regions, and deploy an AI-powered meteorological warning system.
On present evidence, it mostly seems like an institutionalized container for AI-related activities already happening bilaterally, especially with Global South countries, and its vagueness and framing fit China’s initiative-announcement patterns of recent years. It remains to be seen whether the political announcement will translate into concrete projects and outcomes.
WAICO frames AI safety and governance as the distribution of capabilities among states that do not have them, preventing “new historical injustice in AI” and ensuring that AI development is not “a solo performance by a single country, but a symphony of international cooperation.” WAICO’s attractiveness lies precisely in its potential to offer AI capabilities to its member states. And in light of the US’s confusing and rapidly changing policies, such as brief export licenses for foreign access to Anthropic’s models and pre-release reviews for frontier models, China might seem a more stable, predictable, and cooperative partner of the two powers racing to develop AI.
Based on what is currently known about the organization, this grouping may serve China less by directly shaping governance principles than by broadening its partner portfolio, building a standing constituency for its UN positions, and opening channels for Chinese AI deployment. Instead, norms and standards might diffuse with infrastructure and model integration rather than through documents. Chinese open-source models are already becoming increasingly widespread internationally due to their low cost and permissive licensing, which might gradually open pathways to adopting China’s governance and safety principles and contribute to a geopolitical divide in AI.
However, studies have found that some leading Chinese models remain more susceptible to jailbreaks and other safety failures than comparable Western frontier models. If WAICO encourages more rigorous safety testing and the wider adoption of robust safety standards across Chinese (and other) models, it could contribute to a safer global AI ecosystem as their international use continues to expand.
Alicia García Herrero, Chief Economist for Asia Pacific and the Middle East at NATIXIS and Senior Research Fellow at BRUEGEL
At the 2026 World Artificial Intelligence Conference in Shanghai, Xi framed China’s AI ambitions in unmistakably ambitious terms. Opening the event, he urged the world to seize the “historic opportunity” of open-source AI, presented Beijing as a provider of international public goods, and welcomed the new World AI Cooperation Organization headquartered in Shanghai. The message was clear: China intends to dominate the next phase of artificial intelligence not by matching American closed models dollar-for-dollar, but by making open-weight systems the global default.
The logic is straightforward. US frontier labs have built proprietary platforms that command premium pricing and lock users into their ecosystems. Chinese labs—DeepSeek, Moonshot, Alibaba’s Qwen, Z.ai, and others—have instead released increasingly capable models with open weights. Anyone can download, run, fine-tune and deploy them. The result, Xi’s team believes, is rapid global adoption, especially across the Global South, where cost and access matter more than the last percentage point of benchmark performance. By becoming the price-setters and the infrastructure of choice, Chinese models can erode the market share and pricing power of their American rivals. Recent releases such as Moonshot’s Kimi K3 have already shown that Chinese systems can approach frontier quality at a fraction of Western list prices, turning the token itself into a commodity export.
This is a coherent strategy for technological influence and strategic autonomy. It plays to China’s strengths in rapid iteration, manufacturing scale, and state-backed coordination. It also aligns with Beijing’s broader narrative of inclusive development versus Western technological containment.
Yet the central risk is commercial, not technical. The same open-weight approach that drives adoption systematically strips the model layer of pricing power. When anyone can use the weights, no one can charge a durable premium. The result is a peculiar spectacle: world-class technical achievement generating almost no visible profit at the point of production. Extreme competition—what the Chinese call neijuan, or “involution”—keeps compressing margins across the sector. Listed Chinese AI companies that have delivered genuine breakthroughs remain unattractive as investments; a sector can lead the world in capability and still be a poor place in which to seek market gains. Equity markets have already rendered their verdict, with technology shares selling off even as Chinese labs post successive advances, reflecting investor skepticism that capability will translate into sustainable earnings.
China can lead in capability and still struggle to build profitable national champions. If open weights remain the dominant doctrine and involution persists, the companies expected to carry Beijing’s AI ambitions may lack the financial strength to sustain the race. Technical sovereignty is real; commercial viability is the unresolved constraint.
