Key facts
- The UN launched the UN System Data Commons to make its global statistics accessible to AI systems.
- The platform is built on Google's open-source Data Commons and supports the Model Context Protocol (MCP).
- UNICEF's benchmark test found large language models had an average accuracy of 21.2% for global development indicators.
- Visits from AI assistants to UNICEF's data website increased 67% year-over-year.
- Google.org provided $2 million in funding and technical support for the platform.
- The UN aims to have 80% of its statistical datasets on the platform by 2027.
The United Nations announced Thursday it is collaborating with Google to create the UN System Data Commons, a new platform designed to make the UN's extensive global statistics more accessible and usable by artificial intelligence systems. This initiative aims to address the growing reliance on AI tools for data retrieval while tackling concerns about the accuracy and reliability of the information they provide.
The UN System Data Commons is built upon Google's open-source Data Commons platform and incorporates the Model Context Protocol (MCP), which allows AI systems to directly connect with external data sources. This new system replaces the older UNData portal, offering users the ability to search for statistics across various UN agencies using natural-language queries.
Concerns about AI data reliability were highlighted by João Pedro Azevedo, UNICEF's chief statistician. A UNICEF benchmark test involving six large language models (LLMs) found they achieved an average accuracy score of just 21.2% when answering questions about global development indicators. The study, which is a working paper awaiting peer review, also indicated that models often hedged their answers or provided inconsistent numerical responses when re-queried.
UNICEF has observed a significant increase in traffic to its data website from generative AI assistants, with visits from users clicking links in ChatGPT answers rising 67% year-over-year between January 1 and September 14. AI assistants are estimated to account for about one in ten visits to the site, which receives over 6 million visits monthly.
Twenty-six UN entities have committed to the Data Commons, with data from nearly 20 available at launch. The UN aims to integrate 80% of its statistical datasets onto the platform by 2027. Google.org has provided $2 million in funding and technical support for the platform's core infrastructure, with the intention that the UN will eventually operate and scale it independently. Google initially launched Data Commons in 2018, adding MCP support last year to enable AI agents to query statistics and their sources directly.