The takeaway: António Guterres opened the UN's first Global Dialogue on AI Governance this morning with the sharpest speech he has given on AI — "We cannot vibe-code the truth. We cannot vibe-code the future of humanity" — a demand that lethal autonomous weapons be banned by international law, and three named proposals: an AI child-safety pledge, a Global Fund for AI, and mandatory environmental disclosure for the industry. None of it is binding, and the governmental plenary is still speaking as we publish. But as an agenda, it's the most concrete the UN has ever been on AI.

The speech

The Secretary-General opened this morning's session at Palexpo — 193 member states, the first gathering of its kind — with language that made the wire leads write themselves (AFP, Reuters, UN News):

"We cannot vibe-code the truth. We cannot vibe-code the future of humanity."

The framing throughout was speed against consent. AI, he said, is "a technology that can reshape economies, transform the world of work, sway elections and tilt the balance of security" being "deployed faster than anyone — including the people building it — can keep up"; "an experiment is being run on our own societies, without a plan and without consent." On agentic AI — systems "writing code, acting online and making choices with less and less human oversight" — he was blunter than the UN has been before: "Our institutions were built to govern machines that follow commands. They are not ready for machines that decide."

The hardest line was on autonomous weapons: "Let us call them what they are: Killer robots. Machines selecting and engaging their target and taking a life — without human control and judgement. That is morally repugnant… And it must be banned by international law."

And the closing register: "We may be the last generation able to set the terms on which humanity and machines coexist."

The three proposals

Rhetoric aside, the speech carried named asks (UN News, AFP):

1. An AI Child Safety Pledge — safety testing before any deployment to children ("We do not let medicine reach a child until it is proven safe. No child should be a guinea pig for unregulated AI"), zero tolerance for child sexual abuse material, and a mandatory handoff to human support when a system detects a child in distress.
2. A Global Fund for AI — he will urge the General Assembly to create it, "to build skills, data and affordable computing power everywhere", aimed at the divide the scientific panel quantified: "We cannot allow the digital divide to harden into an AI divide and the AI divide to become a development gap, a security gap, and a sovereignty gap."
3. An AI Environmental Transparency Initiative — a reiterated call for major AI firms to disclose their full carbon, water and land footprint, and to run data centres on renewables by 2030.

All three are proposals, not decisions — this Dialogue produces a co-chairs' summary, not rules, and everything Guterres asked for routes through the General Assembly or voluntary industry action. Our morning framing stands. But "machines can inform, but humans must decide, and answer" is now the UN's stated doctrine on AI accountability, said to the face of every government at once.

What's still to come

The governmental plenary — three minutes per country on the published programme, five for heads of state or government — is under way as we publish, and no national statements have been reported from the room yet. The two to watch: the US, which confirmed it would send a delegation (level unannounced) after rejecting "centralized control and global governance of AI" at the UN last September, and China, attending at ministerial level and on record that it wants "the UN as the main channel for global AI governance". Day two runs tomorrow; we'll cover what the plenary actually produces.

Disclosure: On The Wire runs on Anthropic models. We flag it every time we cover the company. Updated coverage of the Dialogue continues in the Daily Update.