Abstract

Artificial intelligence (AI) can support national statistical institutes at several stages of their work: from processing and classifying information to improving statistical production, making data easier to access and helping users understand statistical results. For an institute such as INSEE, the key challenge is to harness these technologies while preserving the principles that underpin official statistics: quality, professional independence, confidentiality, transparency and equal treatment of users.
Insee has introduced a charter governing the use of generative AI to ensure that its adoption is responsible and controlled, alongside staff guidance, awareness-raising and training. Insee’s AI charter applies to all staff and sets out six principles for using the AI assistants currently available: security and confidentiality, digital sustainability, responsibility, transparency, sovereignty and independence, critical thinking. These principles evolve as AI tools develop, particularly as Insee works to provide secure, sovereign tools. Their implementation is supported by practical guidance and training tailored to staff needs. Insee identifies existing training resources and additional needs, and regularly updates its training offer to keep pace with this rapidly changing field. A community of practice enables staff to share use cases and exchange good practices.
Regarding communication and dissemination, several proofs of concepts have been set up in 2025-2026. The most critical one aims at making INSEE’s flagship website “AI-ready”. Another one targets the analysis of verbatim comments by users and by respondents to our surveys. A third one focuses on extracting data from print tables, in order to assist researchers in using historical time series.
The presentation will quickly present and discuss those three proofs of concepts.