UN Partners With Google to Build AI-Ready Data Commons

The United Nations is collaborating with Google to make its global statistics more accessible for artificial intelligence systems. [1]

The new platform, called the UN System Data Commons, is built upon Google’s open-source Data Commons framework. [1] It enables users to search for statistics across UN agencies using natural-language queries, replacing the traditional database interface of the previous UNData portal. [1] The system also supports the Model Context Protocol, which permits AI systems to connect directly to these external data sources. [1]

The transition occurs as users increasingly rely on AI tools, despite current systems struggling to surface authoritative data reliably. [1]

A UNICEF benchmark of six large language models across over 133,000 responses to questions on global development indicators found an average accuracy score of just 21.2 percent. [1] The models tested included OpenAI’s GPT-4o and GPT-4o-mini, Anthropic’s Claude Sonnet 4.5 and Haiku 4.5, and Google’s Gemini 2.5 Flash and Gemini 2.0 Flash. [1] UNICEF also reported that visits to its data website referred from ChatGPT answers increased by 67 percent year-over-year between January 1 and September 14. [1]

The United Nations' adoption of the Model Context Protocol establishes a high-profile use case for how large institutions can explicitly bridge their proprietary data and external AI models. This infrastructure upgrade directly addresses the poor 21.2 percent baseline accuracy seen across leading models from OpenAI, Anthropic, and Google when tasked with querying global development metrics. As AI referrals continue to drive surging traffic to agency sites, structured access should help mitigate hallucinations. What the sources don't address: Whether other major intergovernmental organizations plan to adopt the same Google-backed data architecture for their own portals.