A New Institution, A Familiar Pitch
On July 16, 2026, on the eve of the World AI Conference in Shanghai, representatives from 29 countries — including Russia, Pakistan, Indonesia, Kazakhstan, Laos and Brazil — signed the agreement establishing the World Artificial Intelligence Cooperation Organization (WAICO), the first intergovernmental body dedicated solely to AI (Xinhua/China Daily via gov.cn). China's Foreign Minister Wang Yi signed on Beijing's behalf; UN Secretary-General António Guterres attended. Headquartered in Shanghai, WAICO is pitched as an "independent intergovernmental international organization" committed to "extensive consultation and joint contribution for shared benefit."
President Xi Jinping framed the launch in his keynote the next morning, arguing AI development "should not be a solo performance by any single country but rather a symphony of global cooperation," and explicitly warning against "overstretching the national security concept in the field of AI and placing one country's security over that of others" — an unmistakable jab at Washington's export-control regime (Chinese Foreign Ministry, full keynote). Xi paired the rhetoric with concrete development pledges: 5,000 AI training slots for developing countries over five years, joint application centers with ASEAN, the African Union, the Arab League and BRICS, and access for 30 countries to China's MAZU AI weather-warning system.
The Steelman: A Real Gap in Who Writes the Rules
The grievance WAICO is built on is not manufactured. The EU's AI Act, the G7 Hiroshima process, and the US-anchored patchwork of executive orders and export controls were negotiated almost entirely among wealthy democracies, then presented to the rest of the world as a fait accompli. A recent institutional mapping of 15 global AI governance instruments found the "Western cluster" — the EU AI Act, the Global Partnership on AI, the OECD framework — clusters tightly around values-gated or capability-gated membership, scoring strongly negative on any measure of development orientation; WAICO is the first instrument in that dataset to combine universal, precondition-free membership with an explicitly development-first mandate (arXiv working paper, "Mapping an Emerging Institution in the Global AI Governance Regime Complex"). For a country with no domestic compute, no seat on the OECD AI Policy Observatory, and no leverage over Nvidia's export license queue, a body that asks only for a signature — not a human-rights scorecard — is a rational thing to join. Cheaper, open-weight models genuinely do lower the barrier to entry for the developing world, and a publication built on pro-innovation, open-internet values should welcome more competitive supply in the model layer, not just more compute in Virginia and Loudoun County.
Where the Framing Breaks Down
The trouble is that WAICO's promise of "openness" is not the same openness the People's Republic applies to its own AI sector. The Cyberspace Administration of China's Interim Measures for the Management of Generative AI Services, in force since August 15, 2023, requires that generative outputs adhere to "core socialist values," bars content touching on subversion of the state or ethnic and social division, mandates algorithm registration with the CAC, and layers in security assessments for any service with "public opinion attributes" (CAC, official measure notice). That is the same government now exporting a governance vocabulary of "open source, openness, collaboration and sharing" to 28 co-signatories. The Shanghai Declaration issued alongside WAICO's founding is explicit that data should remain "manageable, controllable and traceable" even as it enables cross-border flows (Chinese Foreign Ministry, Chair's Statement) — a formulation lifted directly from China's own cybersecurity and data-security law architecture, not a neutral technical standard.
That gap matters because WAICO's institutional substance is still, by its own drafters' admission, thin. The arXiv mapping notes formalization remains incomplete: no public charter, no named secretariat, no disclosed funding formula, no decision-making procedure. "Who pays, who decides" remains unresolved. An organization with undefined governance and a founding member whose domestic playbook is content-licensing and algorithm registries is not obviously safer for a small state's sovereignty than the Western club it's positioned against — it is a different sovereign, offering a different set of strings.
What to Watch
Three things will determine whether WAICO becomes more than a diplomatic signal: whether it publishes an actual charter and secretariat structure in the next 12 months; whether any open-weight developer outside China's ecosystem — Mistral, Meta's Llama successors — treats WAICO as a genuine standards venue rather than a Chinese state platform; and whether members beyond the initial 29 actually sign on once the compliance obligations, if any, become clear. Until then, the fairest reading is that WAICO correctly diagnoses a real legitimacy problem in Western AI governance — and answers it with an institution whose founder has not extended the same openness to its own citizens.