THE CRUNCH

The United Nations has launched the UN System Data Commons, a new platform built on Google's open source Data Commons that makes statistics from across UN agencies searchable with natural-language queries and accessible to AI systems. It replaces the older UNData portal, where users had to browse a traditional database interface, and supports the Model Context Protocol (MCP), a standard that lets AI systems connect directly to external data sources. Google.org provided $2 million in funding and technical support, and the system runs on a UN-governed instance intended to be maintained independently by the UN.

The motivation is spelled out by a UNICEF benchmark, described by the agency's chief statistician João Pedro Azevedo in a briefing: six large language models, including OpenAI's GPT-4o, Anthropic's Claude Sonnet 4.5 and Google's Gemini 2.5 Flash, averaged just 21.2% accuracy across more than 133,000 responses to questions about global development indicators. About three in five responses gave no usable number at all, often because the models hedged. The study is a working paper not yet peer-reviewed.

Consistency is another problem. When the same questions were rerun on the same model versions about two days later, models that provided a number both times returned the identical figure only about half the time. Meanwhile, traffic to UNICEF's data website from ChatGPT referrals rose 67% year-over-year between January 1 and September 14, and the agency estimates AI assistants now account for roughly one in 10 visits to the site.

The UN said 26 of its entities have committed to the Data Commons, with data from nearly 20 available at launch, and aims to move 80% of the UN system's statistical datasets onto the platform by 2027. The platform tracks the original source of each statistic so AI-retrieved figures can be traced back to the UN. Google also demonstrated an AI system connected via MCP pulling together multiple indicators, such as HIV infections and life expectancy, to generate dashboards and analysis automatically, though Google's Prem Ramaswami cautioned that a human should always review such outputs before they are cited or published.

WHAT HAPPENS NEXT

The UN aims to have 80% of the UN system's statistical datasets on the platform by 2027, and UNICEF plans to release its benchmark methodology, code and data alongside the paper.