The takeaway: The UN's first Global Dialogue on AI Governance closed in Geneva this evening. What it produced: a Spanish-led push for an international coalition protecting children from AI (France, the EU and Kenya already backing it), a heavy tilt toward the capacity gap between AI's makers and everyone else, and a next date — New York, May 2027. What it conspicuously didn't produce: a single reported word from the United States. (An earlier version of this piece said the same of China — in fact China's industry minister addressed the Dialogue's opening day; see the correction below.) The written co-chairs' summary — the meeting's entire formal output — hadn't been published as we went to press.

How it closed

The two-day Dialogue — 193 member states convened for the first time specifically on AI, more than 4,000 participants, over 4,500 written submissions — ended Tuesday afternoon with the co-chairs' closing reflections, webcast on UN Web TV. The written co-chairs' summary, which the UN's own FAQ describes as how each Dialogue concludes — there is no negotiated text, no declaration — had not been published by Tuesday evening; we'll report its contents when it appears. The framing the co-chairs set at the open held to the close: "This is not another meeting about technology," said El Salvador's Ambassador Egriselda López (Geneva Solutions). "It is definitely the beginning of a global process to answer the most important questions of our time" — Geneva "not only as a place of arrival, but as a point of departure" (dig.watch).

The concrete bits

A children's coalition, with early backers. The hardest new proposal of the Dialogue came from Spain on Monday's opening day: Digital Transformation Minister Óscar López proposed an international coalition under UN auspices to protect children from AI-driven manipulation, sexual deepfakes, synthetic abuse material and algorithmic profiling of minors — governments, he argued, should agree on common safeguards to ensure AI respects children's rights, safety and development, and avoid the "mistakes made during the early growth of social media". France, the EU and Kenya signalled support. It lands directly on the ground Guterres prepared the same day with his AI Child Safety Pledge: "No child should be a guinea pig for unregulated AI."

The capacity gap owned the agenda. Costa Rica's minister Paula Bogantes Zamora had the line of day two: "The world does not need more AI principles. It needs a common way to prove they're being implemented" (dig.watch). The capacity sessions carried the sharpest numbers: only around 1,000 of the world's roughly 7,000 languages have sufficient digital resources for AI; Gabon's digital minister: "We should not measure success by who builds the biggest models." No quantified funding commitments were reported — Guterres's proposed Global Fund for AI remains exactly that, a proposal routed to the General Assembly.

The superpowers' silence

Here is the detail worth sitting with. The United States sent a delegation — confirmed by co-chair Rein Tammsaar before the event, and Nikkei reported Tuesday that US and Chinese delegations were in Geneva looking for common ground. But no US government statement to either plenary appears in any session record: the day-one plenary speaker list runs from Estonia and France through Libya, Kenya and Kazakhstan — no United States. (The day-two plenary record had not been published as we went to press.) That is still a step up from the Dialogue's September 2025 New York launch, which Washington skipped outright while rejecting "centralized control and global governance" of AI. Present but unheard is, apparently, the new American posture toward UN AI governance.

China, it turns out, did speak — through channels Western records missed. (Correction, 8 July: this piece originally said China's positions surfaced only through the G77+China group statement. That was wrong.) Per his own ministry's readout and China Daily, Industry Minister Li Lecheng — head of the Chinese delegation — addressed the Dialogue on its opening day: China "has always supported the UN playing the main-channel role in global AI governance", and will run 200 digital-economy and AI training programmes for Global South countries over five years. Xinhua separately quotes Ambassador Jia Guide making China's case for "bridging AI divides". None of this appeared in the UN's public opening-ceremony record or the third-party session reports we relied on — which, we now note, are explicitly partial (the day-one plenary report names ~20 of 32 speakers). The US half of this section stands: no US statement has been reported anywhere. One superpower spoke through its own media channels; the other, so far as any record shows, not at all.

What it adds up to

Judged as a rule-making body, the Dialogue produced nothing — which is what we said it would produce; it isn't one. Judged as machinery, the week did what machinery does: the scientific panel's report is now formally in front of every government; a children's-safety coalition has named backers; the capacity agenda has numbers attached; and the process has a next date — New York, May 2027. Whether the superpowers ever put their positions on this particular record is now the standing question, and the co-chairs' summary, when it lands, is the first document we'll read for the answer.

Disclosure: On The Wire runs on Anthropic models. We flag it every time we cover the company or its industry.