UN Global Dialogue on AI Governance

About

This was the inaugural session of the UN's standing intergovernmental dialogue on the governance of artificial intelligence, established under the Global Digital Compact adopted in 2024. It gives all member states a shared table for discussing how AI should be governed internationally.

Held in Geneva at the start of the July digital week, it led directly into the AI for Good Global Summit and overlapped the WSIS Forum. The second dialogue is set for New York in May 2027.

What happened

The co-chairs put attendance at roughly 4,000 participants from more than 170 countries. Secretary-General António Guterres warned that AI is being deployed “faster than anyone… can keep up” and set out four priorities: safety (including an AI child-safety pledge), human-rights red lines, capacity (a global network and a forthcoming global fund for AI), and environmental transparency. He also called for a ban on lethal autonomous weapons. The Independent International Scientific Panel on AI then handed over its preliminary report, and two concrete threads ran through the rest of the day: a co-chairs’ summary that, with the panel’s report, feeds the 2027 dialogue, and a voluntary AI Partnership Hub for registering real initiatives.

Day 2 turned from the what to the how, across two thematic clusters: making AI safe, secure and trustworthy, which became the case for interoperability over harmonisation, and human rights, where High Commissioner Volker Türk argued for “smarter, kinder, wiser” over “bigger, faster, better.” The day’s plenary then compared how governance is actually being built across countries and regions.

The closing set the terms for what carries forward: a non-binding co-chairs’ summary rather than a negotiated text, the Secretary-General’s proposed global fund for AI and his child-safety pledge, a new UN AI Resource Hub, and a second dialogue in New York in May 2027.

What we were watching, and how it landed

• A process being born. The format is now set, and it is deliberately light: a non-binding co-chairs’ summary rather than a negotiated text, carrying forward to New York in May 2027. That keeps the table open, and it means the dialogue’s output is a record of views rather than an agreement anyone is bound by. As of 16 July the co-chairs’ summary had not been published.
• Who’s steering it. Co-chairs Egriselda López (El Salvador) and Rein Tammsaar (Estonia) ran it, and the Independent International Scientific Panel on AI handed over its preliminary report on Day 1. Pairing an intergovernmental table with a standing scientific panel is the structural bet of the whole design: the panel supplies the evidence, the dialogue decides what to do with it.
• Divides, not just risk. This held. Capacity and access ran through both days, in the Secretary-General’s proposed global fund for AI, UNESCO’s capacity-building push, the AI Partnership Hub and the UN AI Resource Hub. The instruments are voluntary, so what they deliver is now a question of who registers and who funds them.

Session coverage

SDGCounting summaries of the segments that matter: what was said, who said it, and why it counts for the governance of AI.

The four thematic clusters, the Day-2 governance-initiatives plenary and the closing session are all covered above.

Why it matters for the SDGs

The dialogue’s premise is that AI governance is a development question, not only a safety one, and the two days largely argued it that way. The gap it kept returning to is the one SDG 10 is meant to narrow, and the framing that did the most work was the distinction between two divides: access to the technology, and the capacity to adopt and govern it. Access alone widens the gap, because a country that cannot evaluate a model, or help write the rules it will be held to, is a rule-taker regardless of how much compute reaches it. The instruments on the table (a proposed global fund for AI, UNESCO’s capacity-building work, the AI Partnership Hub and the UN AI Resource Hub) are aimed squarely at that second divide, and all of them are voluntary.

For SDG 17, the structural question is whether a non-binding dialogue plus a standing scientific panel can hold a field this concentrated to account. The case for interoperability over harmonisation, made across Day 2, is the pragmatic version: safety evidence and accountability that travel across borders, rather than one negotiated rulebook nobody signs. The case against is that voluntary machinery is only as good as its uptake. The scientific panel’s “no guarantees” and Volker Türk’s human-rights red lines are what keep the optimism honest, and human oversight that is, in the phrase from the room, a rubber stamp would satisfy the form of all of this while delivering none of it.

Watch & read

The dialogue was webcast and archived on UN Web TV, which makes it the best-documented event of the Geneva week. The co-chairs’ summary, the meeting’s formal output, had not been published as of 16 July.

• Global Dialogue on AI Governance, official site & programme
• UN Web TV, recordings of the sessions (6 and 7 July 2026)
• UN News on the opening
• Independent International Scientific Panel on AI, our coverage of the panel and its preliminary report