What happened
For the first time, the United Nations gave open-source AI a full dedicated day. It sat inside UN Open Source Week (a builders-first week of hackathons, workshops, and panels that follows up the Global Digital Compact), but where the Compact’s other tracks lean on policy, this one deliberately “took a back seat” on the policy agenda to hear from the people actually building the technology. As host Amandeep Gill, the UN Secretary-General’s Envoy on Technology, put it, there are very few rooms at the UN “where those in suits sit next to those in shorts and make things happen.”
The day’s argument, threaded through nearly every session, was that openness lowers the barrier to entry, counters the concentration of AI power, and is the most realistic path to digital sovereignty, but that access alone is not capability. Openness only delivers, speakers repeatedly stressed, when it is paired with investment in skills, compute, and local innovation ecosystems. A Turing-laureate keynote made the technical case for open, sovereign AI; an all-women ministerial panel showed what it looks like on the ground in three developing economies; a robotics panel asked whether the same openness can reach physical AI; and a closing session on autonomous agents mapped what has to be built before anyone can trust them.
Session coverage
SDGCounting summaries of the day’s three headline sessions: what was said, who said it, and why it counts for SDG progress.
The case for open, sovereign AI
The Turing laureate on why funnelling the world’s information through a handful of proprietary models is dangerous, and how a collaborative open model could surpass them.
Read summary → Ministerial panel · Morocco, Jamaica, Sierra LeoneOpen source AI for development
Three ministers on using open source to escape vendor lock-in, build local capability rather than dependency, and share infrastructure across regions.
Read summary → Keynote & panel · Open AgentsGoverning autonomous AI agents
“Intelligence capital,” sovereign AI as a “third way,” and why verification, not just openness, is the precondition for trusting agents that act on your behalf.
Read summary →Also on the programme
- Setting the scene. Cloudera CTO Sergio Gago argued that “open source AI cannot mean simply publishing the weights”; it needs open data formats, interoperability, and “private AI” that brings the model to the data. A German government representative set out a “public money, public code” agenda: the Sovereign Tech Agency, open-source citizen-service and infrastructure-planning AI, and the openCode / openDesk platforms.
- Open robots. A panel spanning Wandercraft, AI² Robotics (Shenzhen), OpenCV founder Gary Bradski, and Shenzhen Robot Valley asked whether open source can do for robotics what it did for software and AI. The recurring answer: the hard problem is deployment and reliability in an unpredictable real world, and openness helps most through shared data, shared infrastructure, and open safety standards, e.g. a national database of AI/robot accidents that every new system must pass in simulation.
- Closing session. USG Gill closed alongside Spain’s AI-supervision agency, Estonia’s Minister of Justice and Digital Affairs, and Uruguay, agreeing that open source is a governance choice, not just a technical one, and the default that keeps states from being vendor-locked. Spain pointed to its own open, Spanish-trained national model as a counterweight to English-dominant commercial systems.
Why it matters for the SDGs
The through-line of the day was the AI divide: the gap SDG 10 is meant to narrow, playing out in who can build, own, and shape AI rather than merely consume it. The concrete proposals were squarely developmental: open models fine-tuned for under-served languages; small models that run locally and cheaply enough to reach a farmer with a question about a crop; shared regional compute and data infrastructure so that not every country has to build its own; and an “open source first” procurement stance to break the licensing traps that drain public budgets. Jamaica’s minister summed up the stakes in a line that could serve as the day’s thesis: open-source AI must build enduring capability, not dependency.
That connects directly to SDG 9 (industry, innovation and infrastructure), SDG 4 (education and skills), SDG 10 (reduced inequalities), and SDG 17 (partnerships and technology transfer), and to the Global Digital Compact’s unfinished capacity-building pillar. The open question the UN left the room with: openness is necessary, but only investment in skills, compute, and local ecosystems turns it into real participation.
Watch & read
- UN Open Source Week 2026, official week site and programme.
- UN Office for Digital and Emerging Technologies (ODET), the host office.
- UN Web TV, recordings of the AI Day sessions (23 June 2026).
SDGCounting’s summaries draw on the UN Web TV recordings of the day. Quotations are lightly edited from an automated transcript and should be read as close paraphrase; names and titles reflect roles at the time of the event.