The United Nations announced Thursday that it is working with Google to make its global statistics easier for AI systems to read and use. The new platform is called the UN System Data Commons.
It is built on Google's open source Data Commons platform. The system replaces the older UNData portal, which relied on a more traditional search and browse setup.
Users can now search UN statistics using plain language questions instead of digging through a database. The platform also supports the Model Context Protocol, a standard that lets AI tools connect directly to outside data sources.
AI Models Struggled With Accuracy
The push comes after a UNICEF study raised concerns about how AI systems handle global data. The agency tested six large language models across more than 133,000 responses.
The models included OpenAI's GPT-4o and GPT-4o-mini, Anthropic's Claude Sonnet 4.5 and Haiku 4.5, and Google's Gemini 2.5 Flash and Gemini 2.0 Flash.
The average accuracy score was just 21.2%, according to UNICEF chief statistician João Pedro Azevedo. About three out of five responses did not give a usable number at all.
Researchers also tested the same questions again two days later. Even when a model gave a number both times, it matched the earlier answer only about half the time.
The study is a working paper that has not yet been peer-reviewed. UNICEF said it plans to release its methodology, code, and data alongside the final paper.
UNICEF has also seen more traffic coming from AI tools. Visits referred by links in ChatGPT answers rose 67% year-over-year between January 1 and September 14.
Those AI referrals made up 6.4% of all sessions on UNICEF's data site this year. The agency estimates AI assistants overall now account for about one in 10 visits to the site.
Rollout Plans and How It Works
So far, 26 UN entities have agreed to join the Data Commons project. Data from close to 20 of them is already available at launch.
The UN's goal is to have 80% of its statistical datasets on the platform by 2027. Shantanu Mukherjee of the UN Statistics Division said the system connects data across UN agencies at a wider scale than before.
Google.org provided $2 million in funding and technical support to help build the platform's core infrastructure. Prem Ramaswami, who leads Google's Data Commons team, said Google used a train-the-trainer approach so UN staff could take over the system.
Google first launched Data Commons in 2018 to organize public datasets into one framework. It added support for the Model Context Protocol last year, letting AI agents query the data directly.
The new UN platform tracks where each statistic comes from. That lets users trace numbers pulled by an AI tool back to the original UN source.
In one demo, Google showed an AI system pulling together UN data on the U.S. President's Emergency Plan for AIDS Relief in Africa. It used stats on HIV infections, AIDS mortality, and life expectancy to build an infographic without a person manually combining the data first.
Ramaswami said giving AI systems accurate data does not guarantee accurate conclusions. He said models can still misread nuance, so a person should review AI outputs before they are published or cited.
