UN System Data Commons: 26 UN agencies commit to make global data AI-ready

The United Nations has a data problem that has nothing to do with collecting numbers and everything to do with connecting them. For decades, agencies across the UN system have gathered some of the most trusted statistics on the planet — on health, poverty, education, water access — only to lock them inside separate databases that rarely speak to each other. On September 17, 2026, the UN and Google announced a fix: the UN System Data Commons, an open, AI-ready platform designed to pull that scattered information into a single searchable resource anyone can query in plain language.

Key takeaways

• The UN System Data Commons unifies siloed UN statistics into one AI-ready knowledge graph, replacing the older UNData portal.
• It runs on Data Commons, Google’s open-source platform, and supports natural-language search plus AI assistant features.
• Google.org contributed $2 million in funding and technical support, while the UN Foundation backed the project’s rollout.
• Twenty-six UN entities have committed to the platform, with data from nearly 20 already live at launch, and the goal is to have 80% of UN system statistical datasets on board by 2027.
• A separate UNICEF benchmark found leading AI chatbots answered global development questions with only 21.2% average accuracy, underlining why authoritative, traceable data matters.

UN System Data Commons: an AI-ready platform for global statistics

At its core, the UN System Data Commons is meant to end the months of manual spreadsheet-wrangling that analysts previously endured before any real research could begin. Built on Data Commons, the open-source framework Google first launched in 2018 to organize public datasets into a common structure, the new platform replaces the older UNData portal, which relied on a more traditional browse-and-search database interface.

Google.org backed the effort with $2 million in capacity-building funding and technical support to help establish the platform’s core infrastructure, working alongside the UN Foundation. Prem Ramaswami, who leads Google’s Data Commons team, said the system is hosted on a UN-governed instance and is intended to eventually be run independently by the UN itself. “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.

Shantanu Mukherjee, acting director of the UN Statistics Division, framed the launch as a leap in scale rather than a minor upgrade. “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,” Mukherjee said, adding that the moment was also about “taking this moment to also make our data AI-ready.”

Turning UN system silos into one connected knowledge graph

The platform’s main achievement is making previously incompatible datasets speak the same language. Instead of analysts manually reconciling mismatched formats, timelines, and geographic boundaries across agencies, the UN System Data Commons automatically integrates those metrics into one interconnected environment — what the UN describes as an AI-ready knowledge graph. That frees researchers to focus on spotting trends and building evidence-based solutions rather than formatting spreadsheets.

Built on open standards like the Model Context Protocol

Interoperability runs through the platform’s design. It supports the Model Context Protocol, or MCP, a standard Google added to Data Commons last year that lets AI systems connect directly to external data sources and pull statistics along with their sourcing. That technical foundation is what allows AI agents to query the UN’s figures autonomously rather than requiring a human to browse for them one dataset at a time.

Natural language search and AI assistants change data access

Anyone can now ask the UN System Data Commons a question in ordinary language and get back relevant figures and interactive charts, without needing to know which agency holds the underlying dataset. That matters because it opens serious statistical work to people who were previously locked out by technical database interfaces — nonprofit program managers, journalists, and policy analysts among them.

How the platform answers plain-language questions

Users can type questions such as how access to clean water in rural areas affects school attendance, how many people gained electricity access over the last decade, or how life expectancy has shifted across regions, and receive an answer instantly. Those who prefer browsing can use the Explore tab to filter by location or by theme, like health or education, while a Blog section turns complex trends — such as UNICEF data on what reduces child poverty — into readable reports.

The launch also pushes AI further into the actual research workflow. Because the system is built to be AI ready, AI agents can autonomously fetch authoritative figures, connect data across different domains, and package everything into charts, infographics, or draft written reports — work that used to take analysts hours of manual searching. In one demonstration, Google had an AI system trace the impact of the U.S. President’s Emergency Plan for AIDS Relief in Africa; the system pulled UN statistics on HIV infections, AIDS mortality, and life expectancy and turned them into an infographic.

Why accurate data matters: the UNICEF AI accuracy test

The push toward AI-ready statistics comes against a backdrop that makes the stakes clear. UNICEF ran a benchmark across six large language models — 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 — testing more than 133,000 responses to questions about global development indicators. The average accuracy score came in at just 21.2%, according to João Pedro Azevedo, UNICEF’s chief statistician, who described the results in a virtual briefing reported by TechCrunch.

About three in five responses failed to provide any usable number at all, often because the models hedged their answers, Azevedo said. Even more striking: when the same questions were run again on the same model versions roughly two days later, models that did return a number both times gave the identical figure only about half the time. UNICEF’s working paper covering the study has not yet been peer-reviewed, but the organization says it plans to release its methodology, code, and data alongside it.

That inconsistency helps explain why demand for reliable, sourced data is rising fast. UNICEF’s own data site, which draws more than 6 million visits a month, saw referral traffic from users clicking ChatGPT answer links jump 67% year-over-year between January 1 and September 14, according to Azevedo. Those AI-driven referrals now account for 6.4% of all sessions, and UNICEF estimates AI assistants overall drive roughly one in 10 visits to the site. Azevedo said tracing data back to its original UN source is increasingly important as more people lean on AI tools to find and interpret information — which is exactly the traceability the UN System Data Commons is designed to preserve.

Google’s own team is careful not to oversell what AI-ready data can guarantee. “Because models can misinterpret nuance, a human should always review the outputs before citing or publishing them,” Ramaswami said.

Data validation and the road to 2027

Every dataset entering the platform is validated by UN system statisticians and technical experts before it goes live, a step meant to keep AI-generated answers grounded in official, trustworthy facts rather than approximations. Twenty-six UN entities have already committed to joining the Data Commons, and data from nearly 20 of them is available right at launch.

The broader rollout is still in progress. Over the coming year, the UN system plans to keep adding datasets from more entities, with a stated goal of covering 80% of UN system statistical datasets by 2027. Readers can explore the platform directly at data.un.org.

FAQ

What is the UN System Data Commons?

It is an open, AI-ready platform uniting global UN statistical datasets into a single searchable resource for easier data access, built on Google’s Data Commons framework and replacing the older UNData portal.

How does the platform help users explore the data?

It enables natural language search, allowing users like journalists and policy analysts to ask questions in plain language and receive data and visualizations instantly, alongside AI assistant tools that automate data retrieval and reporting.

How trustworthy is the data on the platform?

All datasets are validated by UN system statisticians and technical experts to ensure accuracy and trustworthiness, and the platform keeps track of each statistic’s original source so figures can be traced back and verified.

What future expansions are planned for the platform?

The UN system aims to add datasets from more agencies, with 26 entities already committed and nearly 20 live at launch, working toward covering 80% of statistical datasets by 2027.

Article produced with the assistance of artificial intelligence and reviewed by the editorial team.