Global Leaders Wrestle With What 'Open Source AI' Means — and Who Gets Left Out

The second UN Digital Cooperation Day convened ministers, tech executives, and civil society leaders to confront an uncomfortable reality: the world is building AI governance on foundations no one has agreed to define. Across four sessions, a recurring theme emerged — the gap between AI's extraordinary capability and the institutional, governance, and human infrastructure needed to deploy it where it matters most.

• Global panel reveals most AI called "open" doesn't meet the open-source community's own definition, raising concerns about openwashing in international governance
• Brazil questions the trillion-dollar frontier AI arms race, pledging $4.2 billion for sovereign, inclusive AI focused on disaster resilience and digital public infrastructure
• GSMA proposes $100 million fund for cell broadcast early warning systems — mobile networks reach 96% of the world, but only 52 countries have deployed the technology
• Google DeepMind's AI predicted Hurricane Melissa's jump to Category 5 five days in advance, a first for hurricane forecasting
• ODETT and UNDRR launch joint partnership to help the most vulnerable countries absorb AI for disaster preparedness
• Practitioners demonstrate the "30/70 rule": scaling AI requires 70% institutional know-how — data readiness, local ownership, trained personnel — and only 30% technology
• Digital identity system serving nearly 1 million people in Afghanistan and Gaza delivers aid in under 72 hours at 5% overhead, a fraction of traditional costs