Governance

WAICO: China launches a global AI governance body - what it can and can't do

On 16 July 2026, twenty-nine countries signed the founding agreement of the World Artificial Intelligence Cooperation Organization (WAICO) in Shanghai, and on 17 July, Chinese President Xi Jinping formally announced it at the World AI Conference. It is the first standing intergovernmental body built specifically to coordinate AI governance across states - and its founding roster is as telling for who is absent as for who signed. The United States, the European Union, the United Kingdom, Japan and South Korea are not members.

Dated analysis. Written 19 July 2026, two to three days after the signing and Xi's announcement, from reporting by Al Jazeera, Computerworld, CIO, PYMNTS and Chinese state media. WAICO is days old: it has no published charter of company-facing rules, no secretariat track record and no instruments yet. Nothing below reports an obligation your business must meet - because none exists. This is analysis of a governance move, not legal advice.

What actually happened

The sequence is short. Twenty-nine states signed WAICO's establishing agreement on 16 July 2026 at a signing ceremony in Shanghai; Xi announced the organisation the next day and called for "a just and equitable system for global AI governance." The body is headquartered in Shanghai, is described by its founders as guided by United Nations Charter principles, and states a mission to promote "beneficial, safe and fair" AI. UN Secretary-General Antonio Guterres attended the high-level launch event. China's premier, Li Qiang, had first floated the plan a year earlier, in July 2025. (Al Jazeera, Computerworld.)

The founding members skew towards the Global South and China's diplomatic orbit: alongside China, they include Russia, Pakistan, Indonesia, Kazakhstan, Brazil, Malaysia, South Africa and Senegal. The notable gaps are the Western frontier-model economies - the US, the EU and its member states, the UK, Japan and South Korea all sat it out. That absence is the single most important fact about the launch, and I will come back to why.

A cooperation body is not a regulation

Before the geopolitics, a piece of vocabulary that decides how alarmed you should be. An intergovernmental organisation is a body that member states create by treaty to coordinate on something - the WHO for health, the ICAO for civil aviation. It can convene, publish standards, and lend those standards the weight of its members. What it usually cannot do is fine your company. That power stays with national governments, which choose whether and how to turn an international norm into enforceable domestic law.

This is the distinction the launch headlines blur when they call WAICO a "rival to the EU AI Act." The EU AI Act is binding law: if you are in scope and you breach it, the penalty is real - up to the greater of €35 million or 7% of worldwide turnover for the heaviest tier. WAICO, on the current record, is a cooperation body: it sets no company-facing rules and carries no fine. The two things compete for the same prize - influence over the world's default AI rulebook - but only one of them can presently reach into your accounts.

A rulebook and a bloc are not the same instrument

WAICO is a cooperation body, not a regulation. The difference decides whether it is a duty today or a signal about tomorrow.

A regulation (e.g. the EU AI Act)

  • Binding law Obligations attach to named companies in scope.
  • Enforceable penalties Fines are calculated on turnover and ordered by a court or regulator.
  • Defined scope and dates You can tell whether you are covered, and by when.

tell them apart

A cooperation body (WAICO)

  • Intergovernmental forum Member states coordinate; the body does not fine companies.
  • Standards and norms Shapes a default rulebook that members may later adopt at home.
  • Influence, not obligation Its effect sits upstream of the national laws you will eventually meet.

Read WAICO as a signal about tomorrow's standards, not a compliance deadline today.

Two different instrumentsA governance body sets norms that can flow into national law later. A regulation is the law itself. Watching the first is planning; complying with the second is duty - and confusing the two is how a launch headline becomes a false alarm.

Who is out is the real signal

If WAICO cannot yet bind anyone, why does it matter? Because AI governance is fragmenting into blocs, and WAICO is the clearest sign yet that the fragmentation is becoming institutional rather than rhetorical. For a few years the plausible future was a single emerging consensus - some blend of the EU AI Act, the US executive-order approach and voluntary industry commitments - that the rest of the world would broadly track. A 29-state body, headquartered in Shanghai, explicitly courting the Global South and pointedly excluding the Western frontier economies, is a bet on the opposite: competing centres of gravity, each pulling national rules towards its own template.

Three approaches to AI governance, as of 19 July 2026.
ApproachWhat it isDoes it bind companies?
EU AI ActBinding regulation, in force and phasing in through 2027-28Yes - fines up to €35m or 7% of turnover
US federal approachExecutive actions, voluntary commitments, a state-law patchworkPatchy - some state laws bind; no comprehensive federal AI law
WAICOIntergovernmental cooperation body, 29 member statesNo - sets norms, not company obligations

Analysts quoted at the launch read WAICO as a vehicle for Beijing to shape how AI rules are framed at the UN and to win Global South buy-in for a more state-centric model of technology governance. Whether it succeeds is an open question - but the direction of travel for anyone deploying AI across borders is not: expect more than one rulebook, and expect them to diverge.

A worked example: one product, two blocs

Picture a mid-sized company that builds an AI-assisted analytics tool and sells it into both the EU and Indonesia - a WAICO founding member. Today, its binding obligations come almost entirely from where its customers are and what law applies there. In the EU that means the AI Act's transparency duties now and its high-risk regime later; the compliance deadline is a date on a calendar. WAICO adds nothing to that calendar. If the company's lawyer asked "what must we do to comply with WAICO by next quarter?", the honest answer is: nothing - there is nothing to comply with.

Now run the clock forward. Suppose WAICO members converge on a shared norm - say, local registration of frontier models, or data-residency expectations for public-sector AI. WAICO still cannot fine our company. But if Indonesia later writes that norm into its own national law, the company's future obligation in that market traces directly back to a standard being seeded in this body now. That is the mechanism worth understanding: a cooperation body is a norm factory whose output becomes binding only when a member state passes it at home - market by market, on each government's own timetable. The duty always arrives through national law. WAICO just shapes what that law is likely to say.

What to actually do

For almost every business reading this, the correct response to the WAICO launch this week is to note it and keep working. There is no filing, no disclosure and no deadline attached to it. The disciplined move is to separate two questions you should never merge: what am I legally obliged to do today (answered by the EU AI Act, national law, and your contracts) and where are the rules I will face tomorrow being written (increasingly, in more than one place). Track WAICO under the second question. The concrete signals worth watching are whether it stands up a real secretariat, publishes actual standards, and whether any member folds those standards into binding national law. Until then it is a data point about direction, not a task on anyone's compliance plan.

Where this reading could be wrong

Two honest caveats. First, much of the framing around the launch comes from Chinese state media and from analysts reading intent into a two-day-old event; the ambition is documented, the results are not, and new intergovernmental bodies routinely underdeliver on their founding rhetoric for years. Second, the "competing blocs" reading is a forecast, not a fact - it is equally possible that WAICO ends up feeding into UN-level processes that eventually converge with, rather than fracture from, Western frameworks. The safe conclusion is the narrow one: a large, non-Western AI-governance body now exists, the frontier-model economies are not in it, and the era of assuming a single global rulebook is over. What replaces it is still being written.

Questions people ask

What is WAICO? The World Artificial Intelligence Cooperation Organization: an intergovernmental body headquartered in Shanghai. Twenty-nine founding states signed its establishing agreement on 16 July 2026, and Xi Jinping formally announced it at the World AI Conference on 17 July 2026.

Which countries joined WAICO, and who did not? The 29 founding members include China, Russia, Pakistan, Indonesia, Kazakhstan, Brazil, Malaysia, South Africa and Senegal - a coalition weighted toward the Global South. The United States, the EU and its member states, the United Kingdom, Japan and South Korea did not join.

Does WAICO create new rules my company must follow? Not directly, and not yet. WAICO has been presented as a cooperation and standard-setting body, not a regulator; nothing in the founding announcements describes fines or binding obligations on companies. Its influence runs through standards that member states may later write into national law.

Is WAICO a rival to the EU AI Act? They are different kinds of instrument. The EU AI Act is binding law with turnover-based penalties; WAICO is an intergovernmental forum with no enforcement power of its own. Analysts describe it as a China-led alternative for shaping the global default rulebook, especially across the Global South, rather than a competing set of fines.

What should businesses do about WAICO right now? Nothing urgent to comply with. The useful move is to watch it: track whether it produces concrete standards, a working secretariat or model instruments, and factor a multi-bloc AI-governance world into any cross-border AI plan.


Sources & caveats. The date sequence (16 July signing, 17 July announcement), the Shanghai headquarters, the 29-member count, the named members and the absence of the US, EU, UK, Japan and South Korea are corroborated across Al Jazeera, Computerworld, CIO and PYMNTS, with official framing from Xinhua. Guterres's attendance and the analyst commentary are reported by Al Jazeera. WAICO's lack of binding, company-facing rules is a characterisation of a cooperation body days after its founding, not a claim about a published legal text; the EU AI Act's heaviest penalty tier (€35m or 7% of turnover, for prohibited practices) is from that regulation's Article 99. The "competing blocs" and "norms become national law" readings are analysis, not reported fact.