UN Partners With Google to Build Global Data Infrastructure to Ensure Accurate AI-Generated Analysis

The United Nations is rebuilding how its vast trove of global statistics is accessed, turning to Google to make the data easier for AI agents to find, interpret, and use.

The UN has partnered with Google to develop the UN System Data Commons, a centralised platform that brings data from multiple UN agencies into a machine-readable system. A UNICEF working paper found six major AI models correctly answered just 21.2% of over 133,000 questions on global development indicators, highlighting the need for AI to access authoritative data directly.

The UN System Data Commons is being built on Google’s open-source Data Commons platform. It replaces the existing UNData portal and allows users to search across datasets using natural-language queries, rather than navigating individual databases.

The system is designed to make AI agents increasingly act as an interface between users and information. AI systems can directly connect to external data sources and retrieve statistics along with their origin data with the support of the Model Context Protocol (MCP), meaning AI agents can access data from the UN system instead of relying on the model.

“Because models can misinterpret nuance, a human should always review the outputs before citing or publishing them,” said Prem Ramaswami, Head of Google Data Commons, said in a statement.

The UN plans to expand the Data Commons, with 26 UN entities committed and nearly 20 already providing data. It aims to bring 80% of its statistical datasets onto the platform by 2027. Google.org is providing $2 million in funding and technical support.

Data provenance is a key feature that allows AI-generated answers to be traced back to their original UN sources. The platform can also combine multiple datasets through MCP, enabling AI agents to retrieve and analyse information without users manually searching different databases.

Google stated that human review is still required to spot inaccuracies in AI-generated analysis, highlighting the limitations of AI even when using authoritative data. The initiative reflects a wider shift towards making public data structured, machine-readable, and traceable for AI systems.