The Data Innovation and Capacity Branch of UNSD organized a five-session AI Webinar Series in June 2026 for the Global Network of Data Officers and Statisticians. It built on two earlier high-level AI seminars held around the 57th UN Statistical Commission and responded to demand from the 2025 member survey for more webinars on AI and machine learning.
The AI Webinar Series was designed to help the statistical community develop a practical, structured understanding of how artificial intelligence applies to official statistics, moving beyond general discussions to address real-world implications for national statistical systems. It bridged high-level AI-readiness conversations with the practical needs of statistical offices by covering key areas such as data production, privacy-enhancing technologies, governance, and dissemination, while also emphasizing foundational requirements like data quality, interoperability, and institutional capacity.
The series targeted a broad audience - from technical specialists to managers and policymakers - and drew on experiences from countries and international organizations to identify shared lessons, address capacity needs, and foster international cooperation among national statistical offices.
Implementing AI at INSEE: Navigating Between Promise and Complexity
4 June 2026Speakers: Corinne Prost and Conrad Thiounn, INSEE (France)
Webinar 1 offered a candid look at AI adoption inside a national statistical office: at INSEE France, 45% of staff already use AI, but mostly through unofficial "shadow AI" tools rather than institutional channels. INSEE's structural response - two linked innovation teams since 2017, the Onyxia platform (since shared with the UN and now underpinning the UN Global Platform), and a strategic AI committee with real authority over policy, pilot projects, and technology choices, including a deliberate preference for sovereign, open-weight models - illustrates that capturing AI's benefits starts with institutional structures, not just technology.
Privacy-Enhancing Technologies and Artificial Intelligence
9 June 2026Speakers: Matjaž Jug (Statistics Netherlands), Dave Buckley (OpenMined Foundation), Raphaël de Fondeville (Swiss Federal Statistical Office), Benjamin Santos (Statistics Canada)
Webinar 2 took the series into the foundations of trust, with the UN PET Lab's central message that privacy and utility are not a zero-sum trade-off. Tools such as synthetic data, differential privacy, confidential computing, and federated learning let statistical offices extract value from sensitive data precisely by protecting it, as demonstrated through real pilots including secure cross-border record linkage between Statistics Canada and CBS Netherlands.
AI Governance and Data Sovereignty
16 June 2026Speaker: Karla Yee Amezaga, World Economic Forum
Webinar 3 reframed data sovereignty itself: not as geography or isolation, but as meaningful, operational control across the full data and AI life cycle. This represents a shift in governance language from "who owns the data" to "who has which rights, under what conditions" — increasingly relevant as AI agents begin to act on an organization's behalf, with many different actors (states, statistical offices, cities, communities) each holding a piece of sovereignty and a stake in data equity.
Advancing AI Capacity at the CSO for the Production of Official Statistics
23 June 2026Speakers: Brendan O'Dowd, Central Statistics Office (CSO) of Ireland
Webinar 4 grounded the series in practice through Ireland's Central Statistics Office: an honest AI maturity assessment, a new strategy and governance model, and concrete production tools such as cause-of-death and business-activity classification assistants — alongside the CSO's coordination of the 16-country AIML4OS project. A central lesson was that problem definition is the critical first step in any AI use case, particularly for NSOs with limited capacity.
How AI Is Transforming the Dissemination of Official Statistics
30 June 2026Speaker: Marco Marini, International Monetary Fund (IMF)
Webinar 5 closed the loop at the last mile, dissemination. Its central message was that AI-powered access to official data is only as trustworthy as the metadata and standards — such as SDMx (Statistical Data and Metadata) — behind it, and that the central challenge is ensuring what users receive is accurate, traceable, and grounded in authoritative sources rather than convenient approximations. AI is democratizing access to official data, but trust remains the field's defining differentiator.
What's Next
Across all five sessions, one lesson recurred: the technology itself is only one part of this transition; the necessary work that remains is institutional — governance, trust, skills, and international cooperation. The UN Statistical Commission, at its 57th session in March 2026, supported establishing a new City Group on AI-readiness for official data and statistics, led by Rwanda's National Institute of Statistics (NISR) in Kigali. The implementation of AI by statistical offices is also supported by the UN Committee of Experts on Big Data and Data Science (UN-CEBD), which has coordinated the global works in this area since 2014, including the Task Team on PETs behind the UN PET Lab featured in Webinar 2.
Between 2026 and 2028 the Kigali Group is expected to deliver a shared conceptual framework for AI-readiness, the draft MVM, guidance on metadata-driven discoverability aligned with SDMx, guidance on modern dissemination systems and model-friendly endpoints, machine-readable licensing templates, self-assessment scorecards, and a capacity-building package with regional pilots, with particular attention to low- and middle-income countries. An interim report to the Commission's 58th session (2027) will assess progress and advise explicitly on whether an ECOSOC resolution is needed.
The Kigali Group is designed to complement the UNECE High-Level Group for the Modernisation of Official Statistics (HLG-MOS), and will maintain close liaison with UN-CEBD, the SDMx Secretariat, the Committee for the Coordination of Statistical Activities (CCSA), and UN regional commissions — echoing a theme across the series: the community is converging on a small number of coordinated initiatives rather than parallel, duplicative ones.
