Twenty-Nine Signatures in Shanghai: China Builds an AI Governance Body Outside the UN

What happened

On Thursday, July 16, 2026, representatives of 29 countries signed the founding agreement of the World Artificial Intelligence Cooperation Organization — WAICO — in Shanghai, on the eve of the annual World AI Conference. Chinese Foreign Minister Wang Yi signed on behalf of China, according to Xinhua. The organization is headquartered in Shanghai. UN Secretary-General António Guterres attended the ceremony.

The founding agreement states that WAICO "aims to ensure that AI is beneficial, safe and fair, and serves the benefit of all humanity." Al Jazeera describes its goals as promoting international cooperation and developing AI regulation across member states.

President Xi Jinping called the body "an important milestone in the history of AI development" and framed it as "a major move by China to answer the call of the Global South." His most quotable line was a musical metaphor: "AI development should not be a solo performance by a single country, but a symphony of international cooperation." He also warned against "overstretching the national security concept" — a phrase that, in Chinese diplomatic register, is a standing reference to US export controls — and called for "people-centred" governance.

The idea is not new. Premier Li Qiang proposed the organization in July 2025, and Xi returned to it at the APEC leaders' meeting in October 2025. Thursday was the moment it stopped being a proposal.

Who signed — and who did not

Reported founding members include Russia, Kazakhstan, Pakistan, Indonesia, Brazil, Malaysia, South Africa, Senegal, Venezuela, Ethiopia, Cameroon, and Laos. None of the sources fetched for this piece published the full list of 29, which is itself worth noting: a week after the signing, the complete membership of a new intergovernmental body is being reported in fragments.

What is clear is the shape of the roster. It leans heavily on the Global South, plus Russia. Major Western democracies are absent from the charter membership. There is no indication in these reports that the United States, the EU member states, Japan, or the UK were invited-and-declined versus simply not part of the founding cohort — a distinction that matters for how the body is read, and one none of the sources resolve.

The absence is the story's center of gravity. An AI governance organization whose founding members exclude the countries hosting nearly all frontier-scale training compute is not, in any near-term sense, a body that governs frontier AI development. It is a body that organizes everyone else's relationship to it.

Outside the UN, next to the UN

Caixin describes WAICO as an "independent intergovernmental body" operating outside the UN system. That structural choice is deliberate and consequential.

The UN route for AI governance is slow by design — it requires consensus among members with irreconcilable positions on export controls, data sovereignty, and military applications. A standalone body headquartered in Shanghai with a self-selected membership can move faster, set its own agenda, and issue norms without needing Washington or Brussels to agree.

But Guterres's presence complicates the "rival bloc" framing that some coverage has reached for. He described WAICO as a "natural development" of Xi's 2023 Global AI Governance Initiative, said that "technology that will shape the future of humanity must be shaped by all humanity," and praised China's open-source approach — while warning that AI "risks destabilizing global security, widening wealth gaps." That is a UN Secretary-General blessing a non-UN body. Analysts cited by Al Jazeera expect Beijing to use WAICO to shape UN AI policy from the outside and to advance a state-centric governance vision on behalf of the Global South. Those two readings — complement and competitor — are not mutually exclusive, and the ceremony supported both.

The offer attached

Xi paired the launch with concrete commitments, which is where the diplomacy gets substantive. Per Caixin, he pledged 5,000 AI training slots for developing countries over five years, the establishment of international AI cooperation centers with regional blocs including ASEAN, the Arab League, the African Union, CELAC, the SCO, and BRICS, and deployment of China's "Mazu" weather early-warning system across 30 nations.

Read these as the actual product. For a country with no domestic frontier lab and no path to one, the choice being offered is not between Chinese and American AI governance philosophies in the abstract. It is between a package that includes training capacity, regional institutional infrastructure, and a deployed operational system — and a Western posture that has largely consisted of export restrictions and safety frameworks written elsewhere. The pledges are modest in absolute terms; 5,000 training slots across five years and dozens of countries is a rounding error against the scale of the skills gap. But they are specific, and specificity is scarce in this policy area.

Hype versus substance

Several things about WAICO remain genuinely unknown, and the honest version of this story says so.

None of the sources fetched describe WAICO's governance structure, decision-making procedures, voting rules, or what powers — if any — members have ceded to it. No funding model or budget is disclosed. There is no reported information on secretariat leadership or staffing. There is no publicly reported enforcement mechanism, and no reason yet to assume one exists.

That combination — a signed charter, a headquarters, a mission statement, and no visible machinery — describes a large share of new intergovernmental organizations at birth, most of which never acquire the machinery. WAICO may become a real standards body, or it may function primarily as a convening venue and a diplomatic signal. Nothing in the current reporting distinguishes between those futures.

Xinhua's framing — "mankind creates first global intergovernmental AI cooperation organization" — is also worth handling carefully. Whether WAICO is the first such body depends entirely on how narrowly one draws the category, and the claim comes from a Chinese state outlet reporting on a Chinese initiative.

What to watch

Three markers will tell you which way this goes within a year. First, membership drift: does WAICO stay at 29, or do middle powers with hedging instincts — Gulf states, Southeast Asian economies not yet signed, Latin American countries — join? A body that reaches 50 members including a few OECD economies is a different institution.

Second, output: does WAICO publish anything with technical content — model evaluation standards, incident-reporting protocols, compute-access rules — or does it produce declarations? Declarations are cheap.

Third, the response. There is no US or EU reaction in the sources reviewed here. Whether Washington treats WAICO as a nuisance, a negotiating partner, or a thing to be counter-organized against will shape the next round more than anything said in Shanghai.

The takeaway

Twenty-nine countries signed a document creating an AI governance organization headquartered in Shanghai, outside the UN, with the UN Secretary-General in the room endorsing it. The institutional substance behind that signature is, as of now, unverified — no disclosed budget, structure, or enforcement power. What is verified is the alignment: a large bloc of countries without frontier AI capacity has accepted a Chinese-organized framework for talking about AI, and a Chinese-organized set of offers for acquiring some capacity, at a moment when the alternative on the table from the West is mostly restriction. The compute stays where it was. The diplomatic center of gravity moved.