The takeaway: The transcript is out. Xi Jinping's first in-person WAIC keynote — published in full by China's foreign ministry within hours of delivery — pairs four familiar governance principles with three genuinely concrete pledges aimed at the Global South: 5,000 AI training places over five years, a network of AI application centres for six regional and multilateral groupings, and a Chinese weather-warning system rolling out to 30 countries. One line lands squarely on Washington without naming it. All quotes below are our own translation of the official text.

The message, now that we have it

When we covered Xi's decision to keynote WAIC himself, the story was the messenger: a head of state personally fronting what had been a ministers-and-executives event. This morning in Shanghai he delivered, and the Ministry of Foreign Affairs published the full text — title: "Jointly Building a Fair and Equitable Global AI Governance System."

The speech opens seventy years back, at Dartmouth in 1956, and builds to four questions Xi says the international community must answer together: "When machines begin to think, how do humans live with them? When algorithms take part in decisions, how is safety guaranteed? When technology challenges ethics, how does governance keep up? When the divide keeps widening, how is shared benefit achieved?"

The four points — and the line aimed at Washington

The body of the speech is four principles, and most of them rhyme with what Beijing has said before: openness and win-win cooperation (with explicit encouragement of "open-source and openness" — notable in the week a Chinese lab's "open" claims did some travelling); risk awareness and safety; civilisational diversity; and "true multilateralism", giving full play to the UN's role.

Two sentences deserve verbatim treatment. The first is a safety formulation that could have come from a Western lab charter: AI must remain "a trustworthy tool for humanity," with systems of "laws and regulations, technical monitoring, risk early-warning and emergency response" to "ensure AI always remains under human control."

The second is the geopolitical edge: countries should "jointly oppose the practice of over-generalising the concept of national security in AI, and of placing one country's security above the security of others." No country is named. Nobody in the room needed one — the United States has spent this year running export controls, gated model access and security reviews under exactly that banner, a saga we have covered since the Fable 5 export ban.

Three pledges with numbers attached

Rhetoric aside, the speech contains a deliverables list, announced with "I declare" framing:

• 5,000 AI training places for developing countries over the next five years;
• International AI application cooperation centres built for ASEAN, the Arab League, the African Union, CELAC, the Shanghai Cooperation Organisation and the BRICS countries;
• "Mazu," China's intelligent meteorological early-warning system, deployed to 30 countries — named for the Chinese sea goddess who watches over sailors.

That list is the substance of the "AI as international public good" positioning: every item is aimed at the Global South, and none of it requires Western participation.

WAICO, one day old, gets its coronation

The World AI Cooperation Organization — signed into existence by 29 countries only yesterday — is the speech's centrepiece achievement: "a vision from a year ago has become a real outcome," Xi said, adding that it "will become an important milestone in the history of AI development" and calling it a response to "the call of the Global South."

The line built for quotation: "AI development should not be one country's solo, but a symphony of global cooperation."

Domestic proof points came attached: China's "intelligent economy" core industries now exceed one trillion yuan, and this is the opening year of the 15th Five-Year Plan, which he framed as an "opportunity list" for the international community.

What we make of it

Read against the week, the speech completes a deliberate sequence: sign the institution Thursday, crown it Friday, attach concrete goods for the developing world, and position the US approach — unnamed — as the thing the system exists to route around. The same week, Nobel laureates at the Vatican called for a "coordinated slowdown of frontier AI development". Xi's speech is the opposite bet: faster, wider, more institutionalised — with Beijing holding the convening pen.

What to watch next: whether the full WAICO membership list is published, which Western governments respond and how, and whether the six promised AI centres get locations and budgets. Announcements with numbers are checkable — and we will check.