UN Partners with Google to Fix AI's Global Data Failures
The United Nations has partnered with Google to create a platform that will make authoritative statistical data usable by AI systems. The UN System Data Commons aims to address the issue of AI models providing unreliable answers to global development indicators.
A benchmark evaluation by UNICEF found that leading AI models, including those from OpenAI and Google, achieved an average accuracy of just 21.2% across over 133,000 queries. The problem lies in the fact that data lives in fragmented silos across dozens of agencies, making it difficult for AI models to parse and retrieve accurately.
The UN System Data Commons builds on Google's open-source Data Commons technology and introduces the Model Context Protocol (MCP), which creates standardized connections between the data platform and AI systems. The platform will enable natural-language searches for global statistics and allow users to ask questions in plain English rather than navigating dense databases.
The partnership has received $2 million in funding from Google.org, with 26 UN entities committing to the platform. Nearly 20 contributors have provided data at launch, and the goal is to incorporate 80% of the UN's statistical datasets by the end of 2027.