The new UN System Data Commons replaces UNData and opens up the statistical series of twenty UN entities to natural language querying, including the MCP protocol.
The United Nations announced on 17 September 2026 the launch of a platform built on Google's Data Commons technology to replace the former UNData portal. The aim is to make global statistics directly searchable in natural language by software assistants and agents, with support from the Model Context Protocol so that they can access them without going through a human interface. Twenty-six entities of the United Nations system have committed themselves to the project, almost twenty seeing their data available from the outset; The aim is to contribute 80% of the statistical data sets by 2027. Google.org provides $2 million in funding, the platform being designed to be operated independently by the United Nations.
The motivation lies in an embarrassing finding. A UNICEF-led test submitted six major models, including GPT-4o, Claude Sonnet 4.5 and Gemini 2.5 Flash, to 133,000 development indicator questions. Their average accuracy to recover the correct data capped at 21.2%. In other words, the systems that the public is increasingly questioning are mistaken four times out of five on public figures.
There is nothing theoretical about this: ChatGPT traffic to the UNICEF data site increased by 67 per cent over a year, and visits from conversational assistants now account for about one tenth of the total. The platform shall keep track of the source of each data in order to allow verification.
Read the full article on the original source
Go to the source →