The takeaway: This morning in Geneva, for the first time, all 193 UN member states sit down in one room specifically to talk about governing AI. The Global Dialogue on AI Governance opens at 08:30 Geneva time at Palexpo, with the scientific panel report we covered last week formally presented to governments. What it won't produce is rules — the output is a co-chairs' summary, not a treaty. Also today: the Fable 5 credits cliff lands tomorrow, Alibaba's Claude Code ban takes effect Friday, and OpenAI still hasn't answered the 516-token question.

Geneva: 193 countries, one room, no rulebook yet

The UN's first Global Dialogue on AI Governance opens this morning (08:30 Geneva time) and runs today and tomorrow at Geneva's Palexpo — all 193 member states, plus industry, civil society and academia, co-chaired by El Salvador's Ambassador Egriselda López and Estonia's Ambassador Rein Tammsaar.

The published day-one programme is substantial. Secretary-General António Guterres opens alongside General Assembly President Annalena Baerbock, the ITU's Doreen Bogdan-Martin and UNESCO's Khaled El-Enany. Then Yoshua Bengio and Maria Ressa formally present the Independent International Scientific Panel's first report — the 1 July assessment we led with last week, with its findings that a billion people now use conversational AI weekly, that the US holds roughly three-quarters of the world's top-500 AI supercomputing capacity, and panel co-chair Yoshua Bengio's warning that "science currently cannot guarantee" AI won't cause catastrophic harm. After that: a plenary of governments — three minutes each, five for heads of state or government.

Update, 09:30 BST: The opening ceremony has concluded; official transcripts haven't landed yet (we'll add Guterres's words when they do). Two sourced developments on the questions we flagged this morning: Estonia's co-chair Ambassador Rein Tammsaar confirmed last week that the US administration "will also send a delegation" (Geneva Solutions, this morning) — delegation level still unannounced; and China is attending at ministerial level, having said it supports "the UN as the main channel for global AI governance" (Foreign Ministry, 18 June). Scale, per Geneva Solutions: more than 2,600 participants and over 4,500 written submissions.

Two details worth watching. First, the afternoon panels put Alibaba Cloud's Jian Wang on a UN stage — the same week Alibaba's ban on Anthropic's Claude Code takes effect. Google's Yossi Matias is the other big-tech name; no OpenAI, Anthropic or Meta executives appear on the published programme. Second, the US governmental posture is an open question: no US government official appears on day one's published programme — the closest thing to an official American presence in the line-up is a business association, USCIB president Whitney Baird — and we found no announced US delegation level, nor any boycott. We'll report what actually happens rather than guess.

Update, 14:15 BST — from the plenary: The governmental plenary has now run, and the first from-the-room reporting is in (UN News, Reuters). Guterres opened with the line of the day — "We cannot vibe-code the future of humanity" — plus a killer-robots ban demand and three named proposals; our full piece on the speech is here. The first concrete multi-country move: more than 20 countries backed a UN Global Network for AI Capacity Building. Assembly President Baerbock put a number on AI's "sinister" side: "99 per cent of deepfakes are sexual in nature and 96 per cent target women and girls." From the floor, Libya's Mohamed al-Menfi: "AI cannot be a legitimate resource if African countries cannot make use of it"; Georgia's president warned of AI as an "instrument of totalitarian control and new digital tyranny." China's head of delegation is confirmed as Li Lecheng, Minister of Industry and Information Technology — his remarks, like any US statement, are not yet reported; we'll carry both in tomorrow's day-2 coverage.

Honest framing matters here: the Dialogue is a political forum, not a regulator. It exists because last August's General Assembly resolution created exactly two mechanisms — the scientific panel and this dialogue — and its formal output is a co-chairs' summary. Nobody is signing binding AI rules in Geneva this week. The story is that the machinery now exists, that the science is formally in front of governments, and what 193 states choose to say on the record. The second session is already booked: New York, May 2027.

(Also in Geneva this week, and easy to conflate: the ITU's AI for Good Summit starts tomorrow, and the new AI for Good Global Commission — Marc Benioff and Paul Kagame co-chairing, with Jensen Huang and Anthropic's Jack Clark among members — holds its first meeting Wednesday. Different bodies, same city.)

The briefs

• The Fable 5 credits cliff is tomorrow. Per Anthropic's redeployment plan — unchanged as of this morning — bundled subscriber access ends tomorrow, Tuesday 7 July; from Wednesday 8 July, Fable use draws on metered usage credits. Yesterday we covered what the deadline looks like from the ground: a veteran developer racing to spend his allowance, receipt published.
• Alibaba's Claude Code ban went mainstream over the weekend (TechCrunch, Saturday): effective Friday 10 July, the tool is classed "high-risk software" and staff move to Alibaba's in-house Qoder. Anthropic's account — that the disputed detection code was an anti-reseller experiment, removed on 1 July — is on the record; the ban proceeds anyway.
• Still no answer on the 516 tokens. The GPT-5.5 reasoning-token clustering report we covered Sunday remains open and unanswered by OpenAI — nine days since filing.
• No new models shipped this weekend — our sweep of every major lab came back empty. The live watches: DeepSeek V4 (official release due mid-July), GPT-5.6's public launch, and Gemini 3.5 Pro's slipped July target.

We'll fold in the Geneva opening session as reporting lands through the morning. Disclosure: On The Wire runs on Anthropic models. We flag it every time we cover the company.