WAICO: China gives itself a world AI organization — 29 countries, and still not one published rule

On 16 July 2026 in Shanghai, under the eyes of the United Nations Secretary-General, 29 countries signed the founding act of a new intergovernmental organization: the World AI Cooperation Organization (WAICO). Driven by Xi Jinping and presented as “an important step in the history of AI development”, it aims to become a global hub for artificial intelligence governance — led by China, with no Western democracy at the table. Le Recul checked what that solemn announcement actually covers: a timeline, a list of signatories, and above all the question that matters — what does this organization have the power to do, today?

What was signed, and how long the idea has been around

The idea was not born on 16 July. Xi Jinping had already launched a first Global AI Governance Initiative back in 2023. In July 2025, Premier Li Qiang extended it with a Global AI Governance Action Plan, a thirteen-point roadmap covering AI safety, infrastructure, data standards and sustainable development. WAICO is its institutional culmination: signed in Shanghai on 16 July 2026, the day before the World AI Conference (WAIC) 2026 opened, by Chinese foreign minister Wang Yi and representatives of at least 28 other countries — Kazakhstan, Laos, Pakistan, Russia and Indonesia are among the confirmed official signatories. Xi Jinping used the platform to promise 5,000 AI training places to developing countries.

It was a busy week at every level: the very next day, the opening of WAIC 2026 served as a showcase for more than 300 new products, including the launch of Kimi K3 by the Chinese startup Moonshot AI — a model we later put head to head with the most expensive model on the market. Displayed governance and a show of technological force moved, that week, in step.

The 29 founding countries

The full list, as published: Algeria, Belarus, Brazil, Cambodia, Cameroon, China, Congo, Cuba, Ethiopia, Indonesia, Kazakhstan, Kenya, Kyrgyzstan, Laos, Lesotho, Malaysia, Mozambique, Myanmar, Nicaragua, Oman, Pakistan, Russia, Senegal, Serbia, South Africa, Tajikistan, Uzbekistan, Venezuela and Zambia — roughly 10 African countries, 12 Asian, 4 Latin American and 3 European outside the European Union.

One point deserves stating without bending it either way: no Western democracy, no G7 member, has signed. But the list is not simply a club of authoritarian regimes either — Brazil, Indonesia, Malaysia and South Africa are functioning democracies and G20 powers, which made a choice of non-alignment here rather than an ideological endorsement.

An organization… to do what, exactly?

This is where the gap between stated ambition and institutional reality becomes stark. An academic analysis published on arXiv, which rates international organizations on whether they have a public charter, a secretariat, a budget, voting rules and a defined membership list, places WAICO at the bottom of its ranking: it has a name, a host city and a declaration of intent, but the machinery that would let it act has not been specified. A more detailed institutional analysis (Remio) lists seven elementary questions — voting rules, financial commitments, membership conditions, sanction powers, articulation with existing organizations, full membership list, ratification timetable — and finds that not one has a public answer to date. A contributor to Nature, quoted in that same analysis, goes further: WAICO will in all likelihood not govern the AI industry in any binding way.

The founding agreement itself stays cautious on substance: it speaks of AI development that is “beneficial, safe, fair, healthy and orderly”, without specifying whether the organization will one day be able to set standards, run evaluations, certify systems, or merely issue recommendations. No date has been communicated for a first summit, for standing up an operational secretariat, or for a first working session — WAICO exists, for now, on the paper of a signed agreement, not in a calendar of activity.

An already very crowded landscape

WAICO is not landing on empty ground. It joins an already dense stack of initiatives: the OECD AI principles (2019, updated in 2024, 47 adhering jurisdictions), the G7 Hiroshima Process on AI (2023), the EU AI Act — whose core provisions apply from 2 August 2026, and whose machine-readable marking obligation we watched land on a real product when we compared Kling 3.0 and Veo 3.1 — the American AI Action Plan, and the UN’s own dialogue on the subject: the first Global Dialogue on AI Governance, opened in Geneva on 6 and 7 July 2026, backed by the preliminary report of an international scientific panel co-chaired by Yoshua Bengio.

One summary of that accumulation captures the problem well: the world is short of neither AI principles, nor national laws, nor diplomatic forums, nor a UN dialogue — what is missing is agreement on who gets to set the rules. Nothing in the WAICO agreement indicates, at this stage, how it intends to overcome that fragmentation rather than add another layer to it.

The urgency is not merely diplomatic. The preliminary report of the UN scientific panel, co-chaired by Yoshua Bengio, finds that AI capabilities are advancing faster than science itself can measure them: on Humanity’s Last Exam, a test designed to be hard, the best score obtained by an AI went from 8 % to 45 % in sixteen months. More than a billion people now use a conversational AI every week, and agents able to act with growing autonomy — browsing the web, writing code, coordinating other AIs — are becoming common. That pace is what feeds WAICO, the Geneva Global Dialogue, and much of the regulation currently under discussion, all at once.

The paradox: governance abroad, control at home

This is the criticism most repeated in the West, and it deserves stating without caricature. France was among the most explicit: French officials warned that WAICO risked “normalising authoritarian AI governance models”, and described it as an attempt to weaken the Hiroshima Process. The American think tank Carnegie Endowment for International Peace reaches a neighbouring conclusion in a more analytical register: China “approaches AI governance through the lens of strengthening state control” — a reading consistent with Beijing’s domestic use of AI (widespread facial recognition, the social credit system, mass surveillance), and hard to square with the “ethical”, “human-centred” governance language used internationally. Neither camp has a monopoly on consistency here, as the rest of this shows.

This is not a first for Chinese diplomacy at the UN: as early as 2024, the General Assembly had already adopted by consensus, with China’s active support, a first resolution on AI capacity building, carried by a “Group of Friends” gathering more than 80 countries and UN entities. WAICO extends a diplomatic strategy that has been running for several years, not an initiative that appeared out of nowhere on 16 July.

The unexpected (and ambiguous) UN backing

One fact surprises, and complicates the story of a simple China-versus-the-West standoff: UN Secretary-General António Guterres personally welcomed WAICO in his opening address at WAIC on 17 July 2026: “President Xi’s launch, in 2023, of the Global AI Governance Initiative was a strong message of Chinese commitment and leadership, and WAICO is its natural development.”

That statement needs qualifying on two counts: it commits Guterres personally, not a UN resolution or institutional mandate; and the organization is running, at the very same moment, its own competing process — the Geneva Global Dialogue on AI Governance and the scientific panel co-chaired by Bengio, launched barely days before WAICO. Both efforts present themselves as complementary. Nothing at this stage guarantees they will not tread on each other — all the more so since UN-China cooperation on AI did not start this week: more than 20 countries, China included, have already attached centres to a UN-backed Global Network for AI Capacity Building that predates WAICO.

Washington is not joining, but is playing the same tune

The United States has not joined WAICO — no surprise, since the Trump administration’s AI Action Plan mentions international AI diplomacy only to “counter Chinese influence in international governance bodies”, with no positive agenda of its own, against a backdrop of budget cuts hitting the State Department’s cyber diplomacy.

But Washington is not sitting still: back in December 2025, the State Department launched Pax Silica, led by under secretary Jacob Helberg, a coalition focused on securing supply chains for critical minerals, semiconductors and AI infrastructure. Expanded at a summit in June 2026, it now counts 24 signatories: Australia, Finland, India, Israel, Japan, Norway, Qatar, South Korea, Singapore, Sweden, the Philippines, the United Arab Emirates, the United Kingdom and the United States, since joined by Argentina, Chile, Costa Rica, El Salvador, the European Union, Germany, Greece, Kazakhstan, the Netherlands and Panama. Taiwan, without being a full member given its status, separately endorsed the Pax Silica principles through a joint declaration with Washington — a signal that reads all the more clearly given the island concentrates most of the world’s production of the most advanced semiconductors.

The parallel is striking: both camps tie access to their AI technology and infrastructure to diplomatic alignment, and neither has so far published clear binding rules. One detail shows this is not a hermetic bloc against another: Kazakhstan has signed both WAICO and Pax Silica — a way of not picking a side for as long as neither has demonstrated real power.

What to remember

On 16 July 2026, 29 countries — not one of them a Western democracy — signed the creation of WAICO, an AI governance organization based in Shanghai and driven by Xi Jinping, continuing a Chinese initiative launched back in 2023.

On seven basic institutional criteria (voting rules, funding, membership, sanctions, articulation with the existing order, membership list, timetable), not one has a public answer to date according to the available analyses.

France warned that WAICO risked “normalising” authoritarian AI governance; the UN Secretary-General personally welcomed it as the natural extension of Chinese AI diplomacy since 2023 — two readings that coexist without reconciling.

The United States has not joined WAICO, but has been building its own alliance since December 2025, Pax Silica (24 signatories), on the same mechanics: tying technological access to diplomatic allegiance.

The number to remember

0.

That is, according to the most detailed institutional analysis available to date, the number of voting rules, financial commitments or sanction mechanisms WAICO has made public — out of the seven criteria that usually define an international organization able to act. Twenty-nine countries signed a name, a host city and a declaration of intent. What this organization will have the power to do remains, for now, to be written.