United Nations partners with Google to launch AI-ready global data platform

The United Nations and Google launched the UN System Data Commons, a platform built on Google’s open-source Data Commons to let AI agents and users query UN statistics using natural language.

The new platform replaces the older UNData portal, which required users to browse and search through a traditional database interface rather than using natural-language queries.

The UN System Data Commons also supports the Model Context Protocol (MCP), enabling AI systems to connect directly to the UN’s data sources for retrieval and integration.

A UNICEF benchmark of six large language models—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—produced an average accuracy of 21.2% on more than 133,000 responses to questions about global development indicators.

About three in five responses did not provide a usable number, often because models hedged their answers, and when models did provide a number, identical figures were returned only about half the time in repeat tests conducted roughly two days apart.

The UNICEF study is a working paper not yet peer-reviewed, with plans to release methodology, code, and data alongside publication.

Google.org provided $2 million in capacity-building funding and technical support to establish the platform’s core infrastructure, with the UN aiming to bring 80% of its statistical datasets onto the platform by 2027.

Twenty-six UN entities have committed to the Data Commons at launch, with data from nearly 20 entities available initially.

The platform maintains provenance tracking so users can trace statistics retrieved by AI systems back to the original UN sources, addressing concerns about data reliability.

Google demonstrated how an AI system connected via MCP could pull multiple indicators to generate dashboards, charts, and written analysis without manual dataset retrieval.