Post

Shanghai hosts the AI summit that created a global body

The 2026 World AI Conference opened in Shanghai on 17 July, with the signing of the charter for a new World AI Cooperation Organization, a multilateral body aimed at the global south. The summit brought together heads of state and international officials, and its outcome is a bet that governance can be built before the technology is fully understood.

The 2026 World AI Conference opened in Shanghai on 17 July, and its most durable product was signed before the plenary sessions began: the charter of the World AI Cooperation Organization, a new multilateral body born in Shanghai with the backing of the United Nations’ secretary-general and an unusually wide list of governments. The conference itself was the usual mix of showcases and panels. The organisation, if it takes hold, is the part with consequences, because it is an attempt to build governance for artificial intelligence at a scale that has so far eluded every other effort.

The framing from the Chinese side was explicit. The summit’s theme tied development to governance, and the opening address laid out four principles: openness and shared development, safety under human control, mutual respect between civilisations, and genuine multilateralism. The novelty was not in the principles, which echo every AI governance statement of the past three years, but in the delivery mechanism. China announced concrete programmes alongside the charter: five thousand training places for developing countries over five years, cooperation centres linked to regional blocs including ASEAN, the Arab League, the African Union and the Shanghai Cooperation Organisation, and the deployment of an AI-driven weather warning system in thirty countries.

The audience mattered as much as the content. The conference brought together the UN secretary-general, heads of state from Central Asia and Southeast Asia, and ministers from dozens of countries, while the high-level session on governance drew representatives from about eighty states and international organisations. That is a different coalition from the ones that produced the other major AI governance documents of recent years, which were largely Western affairs. The Shanghai body is explicitly aimed at the global south, and the gap between the two coalitions is now the central fact of AI geopolitics: two parallel governance tracks, with different memberships, different priorities, and different ideas about who should set the rules.

It is worth being careful about what the event did and did not accomplish. A charter signing is not a functioning organisation, and the history of multilateral bodies is littered with charters that never quite acquired staff, budgets or enforcement power. The practical test will come with the first programmes: whether the training places are actually filled, whether the weather system is actually deployed, and whether the cooperation centres become real institutions or plaques. The reasonable reading is that this is an infrastructure bet, made early and at modest cost, in the same spirit as the institutions built after the Second World War, on the theory that the rules of a transformative technology are easier to shape before the technology settles than after.

The timing also deserves a note. The summit landed in a year when the frontier of AI capability is moving faster than any governance process, when the United States is pursuing its own national legislative framework, and when the European Union is postponing the enforcement of its flagship law. In that context, a third track, organized by the world’s other AI superpower and pitched at the countries that feel excluded from the first two, is not an academic exercise. Whether it becomes the global south’s venue for AI governance, or a diplomatic gesture that fades, depends on execution over the next few years.

The honest summary is that Shanghai produced an institution rather than a law, and institutions are slower than laws but harder to reverse. The World AI Cooperation Organization has a charter, a host government willing to fund it, and a constituency. The questions it cannot answer yet are the ones every such body faces: whether its members will treat it as a forum or a commitment, and whether the technology it is meant to govern will wait for it. Those answers will come with time, which, for a body founded at this stage of the AI cycle, is both the opportunity and the risk.

This post is licensed under CC BY 4.0 by the author.