The new UN System Data Commons replaces the legacy UNData portal, offering natural-language search and source-traceable statistics from multiple agencies, as a UNICEF benchmark found major AI models frequently get even basic development figures wrong.
The United Nations has teamed up with Google to make its enormous store of global statistics easier to search and use, for both people and artificial intelligence systems, Google said in an announcement on Thursday, 17 September.
The partnership will produce the UN System Data Commons, a single platform that brings together figures from multiple UN agencies in a format built for AI. It runs on Google's open-source Data Commons technology, lets users ask questions in everyday language, and supports the Model Context Protocol (MCP), a standard that allows AI agents to connect to outside data sources. The system replaces the older UNData portal, which relied mainly on traditional database searches. The UN says the change will help AI tools retrieve authoritative statistics while letting users trace every number back to its origin.
A broad coalition and a bold target
Twenty-six UN entities have pledged support for the platform, and data from nearly 20 agencies is available at launch. The organisation aims to have 80 percent of the statistical datasets held across the UN system on the platform by 2027, consolidating most of its statistical output in one place.
The timing reflects a wider shift, as more people turn to AI tools when looking for information. Yet those tools can stumble on basic facts. A UNICEF benchmark covering six AI models, among them systems from OpenAI, Anthropic and Google, analysed about 133,000 responses and recorded an average accuracy of just 21.2 percent on basic development statistics, according to the agency's chief statistician, João Pedro Azevedo. The research remains a working paper and has not yet been peer-reviewed.
From raw data to charts and dashboards
Google's Data Commons team showed how an AI system linked through MCP could generate graphs, dashboards and written analysis drawn from several UN indicators. In one demonstration, the system examined the effects of the US President's Emergency Plan for AIDS Relief in Africa, combining measures such as HIV infections, AIDS-related deaths and life expectancy into an infographic. Google added MCP support to Data Commons last year.
The new platform also shows the origin of each individual statistic, allowing users to inspect the underlying UN data sources. The feature is intended to help people verify answers produced by artificial intelligence. To support the infrastructure behind the platform, Google.org has contributed $2 million in capacity-building grants and technical assistance.
For the UN, the message is that easier access is not enough on its own, since users must also be able to check where a number comes from. By pairing a unified data framework with source tracing, the organisation is betting that AI-generated answers about global development can become more dependable.
