The United Nations unveiled a new partnership with Google on Thursday aimed at making its extensive collection of global statistics more accessible to artificial intelligence systems. The initiative, dubbed the UN System Data Commons, leverages Google's open source Data Commons platform to enable users to search for statistics from various UN agencies through natural-language queries. This new system replaces the previous UNData portal, which required users to navigate a more conventional database interface to find information.

The platform also incorporates support for the Model Context Protocol (MCP), a standard designed to let AI systems connect directly to external data sources. This development comes as more people turn to AI tools for answers, yet many systems continue to struggle with reliably presenting authoritative data. A UNICEF benchmark test involving six large language models and over 133,000 responses to questions about global development indicators yielded an average accuracy score of merely 21.2%, according to João Pedro Azevedo, the agency's chief statistician, who spoke to reporters during a virtual briefing.

The evaluation covered OpenAI's GPT-4o and GPT-4o-mini, Anthropic's Claude Sonnet 4.5 and Haiku 4.5, as well as Google's Gemini 2.5 Flash and Gemini 2. Approximately three out of every five responses failed to provide a usable number at all, frequently because the models gave hedged answers, Azevedo noted. Even more concerning, when the same questions were posed to the same model versions roughly two days later, models that supplied a number on both occasions returned the identical figure only about half the time. This study is a UNICEF working paper being readied for journal submission and has not yet undergone peer review. The organization indicated it will release its methodology, code, and data together with the paper.

UNICEF has also observed a significant surge this year in traffic from generative AI assistants visiting its data website, which garners more than 6 million visits each month and ranks among the agency's most visited sites. Such referrals made up 6.4% of all sessions this year, and UNICEF estimates that AI assistants collectively now represent roughly one in ten visits. The UN reported that 26 of its entities have committed to the Data Commons, with data from nearly 20 available at launch. Furthermore, the organization aims to bring 80% of the UN system's statistical datasets onto the platform by 2027.

"We are orders of magnitude more advanced in scale, scope, and flexibility, connecting for the first time across so many agencies across the UN system," said Shantanu Mukherjee, acting director of the UN Statistics Division. "And [we are] taking this moment to also make our data AI-ready."

Google.org contributed $2 million in capacity-building funding and technical support to establish the platform's core infrastructure. "We have taken a 'train-the-trainer' approach throughout the rollout, and we have already seen the UN system team ramp up quickly," Ramaswami said. Google originally launched Data Commons in 2018 as an effort to organize public datasets from various sources into a unified framework. Last year, it added support for MCP, enabling AI agents to directly query Data Commons for statistics and their sources. The UN's platform also tracks the origin of each statistic, allowing users to trace data retrieved by an AI system back to the original UN source. Azevedo emphasized to reporters that this traceability is crucial as more people depend on AI tools to locate and interpret information.

Beyond enabling AI agents to retrieve individual statistics, Google demonstrated how an AI system connected to the UN data through MCP could aggregate multiple indicators and use them to generate dashboards, charts, and written analysis without requiring a user to manually locate and combine the underlying datasets. In one demonstration, Google asked an AI system to determine the impact of the U.S. President's Emergency Plan for AIDS Relief in Africa. The system identified relevant UN statistics on metrics such as HIV infections, AIDS mortality, and life expectancy, and used them to create an infographic. However, providing an AI system with authoritative data does not automatically make its conclusions authoritative. "Because models can misinterpret nuance, a human should always review the outputs before citing or publishing them," Ramaswami said.