Every AI Policy Ever · open in the interactive register

Published (non-binding, current)Non-binding · Institutional mandate

National Institute for the Evaluation and Security of Artificial Intelligence (INESIA)

Institut national pour l'évaluation et la sécurité de l'intelligence artificielle (INESIA)

Record FR-01

Issued by
Secrétariat général de la défense et de la sécurité nationale (SGDSN) and Direction générale des Entreprises (DGE)
Jurisdiction
France
Level · region
National · Europe
Published / adopted
2025-01-31
Last updated
2026-02-12
Language
French

Summary

France's national institute for AI evaluation and security, created on 31 January 2025 ahead of the Paris AI Action Summit and steered by SGDSN and the Directorate General for Enterprise (DGE). It brings together ANSSI, Inria, LNE and PEReN to build a public capacity to evaluate advanced AI. Its 2026-2027 roadmap sets eleven projects: evaluation tools for AI Act market surveillance authorities, standards work and synthetic-content detection; research on systemic risks from information manipulation, offensive cyber capabilities and CBRN proliferation, and on agentic systems; and performance challenges. SGDSN represents France in the international network of AI safety institutes.

Language

English translation by Secrétariat général de la défense et de la sécurité nationale (SGDSN) Open it. The French text is authoritative.

Safety topics

safety instituteevaluation & testingcatastrophic riskinternational cooperationnational security & militarygeneral-purpose / frontier AI

Status notes

SGDSN announced on 12 Feb 2026 that INESIA had adopted its 2026-2027 roadmap, with eleven projects in three poles (support for AI regulation, systemic risks, performance and reliability) and a cross-cutting axis. SGDSN says the network of AI Safety Institutes was renamed the 'International Network for Advanced AI Measurement, Evaluation and Science' in December 2025. The DGE leads exchanges with the European Commission, in particular its AI Office.

Sources

Open the official source https://www.sgdsn.gouv.fr/nos-missions/anticiper-et-prevenir/evaluer-et-securiser-lia

Secondary source: https://www.sgdsn.gouv.fr/publications/linstitut-national-pour-levaluation-et-la-securite-de-lia-adopte-sa-feuille-de-route

Local copy

Archived. The secondary source returned a bot-check or empty page this time (The page has very little readable text (511 characters); it probably needs JavaScript); the earlier copy is kept. Saved with a headless browser because the plain download looked incomplete. A source changed since the previous snapshot; both versions are kept. Last attempt 2026-10-17 20:41:33 UTC

Official source (third-party copy, not published) archive/FR-01/2026-10-17/official.html
Retrieved 2026-10-17 20:41:37 UTC
SHA-256 12e94c1a2b47e373fc104388860767418a9a73636d0c484ea7d08fa9676849fc

Warning: The page has very little readable text (511 characters); it probably needs JavaScript

Official source, rendered in a browser (third-party copy, not published) archive/FR-01/2026-10-17/official-rendered.html
Retrieved 2026-10-17 19:20:26 UTC
SHA-256 cfcca91027a48a7d947b17ebfd2160def0fd81d6a36256edb20113597a39cef7
Official source, printed to PDF by a browser (third-party copy, not published) archive/FR-01/2026-10-17/official-print-2.pdf
Retrieved 2026-10-17 20:42:52 UTC
SHA-256 0277ed6906c435792de72789bbac4faf6cc77ac4410d923f98e8b709e156f4ec
Secondary source (third-party copy, not published) archive/FR-01/2026-10-11/secondary.html
Retrieved 2026-10-11 14:00:47 UTC
SHA-256 0fefb21d202bd1191ed4175b023a352eeb39219be7ced173d507455595d74adb

Verification

A
The instrument's own official text, or an official register, was stored and confirms the title, date and status. checked 2026-10-17 · official-url-fetched
Stored and read in this run: the SGDSN mission page 'Evaluer et sécuriser l'IA' (dated 5 Feb 2026), rendered in a browser and printed to PDF on 17 Oct 2026, which gives the institute's name, creation on 31 Jan 2025, steering by SGDSN and DGE, members, poles and the network renaming; the INESIA 2026-2027 roadmap (French PDF, sgdsn.gouv.fr), which sets out the eleven projects; the SGDSN news item of 12 Feb 2026 on the roadmap's adoption (stored as the secondary source); press release no. 373 of 12 Feb 2026 (sgdsn.gouv.fr PDF); and SGDSN's English courtesy translation of the roadmap (sgdsn.gouv.fr PDF), which gives the English name of the institute and states that the French document is the official version. No founding decree or standalone founding press release is among the stored files.
Second check: Independent check by Claude Code (claude-sonnet-5-5), 2026-10-17: 6 field(s) supported.

History

  1. 2026-10-17english_version: – → {"kind": "government translation", "url": "https://www.sgdsn.gouv.fr/files/files — the record had none; SGDSN publishes an English courtesy translation of the 2026-2027 roadmap, stored in this run, which says the French document is the official version, so the kind is government translation2026-10-17-recheck-europe-2 · Claude Code (claude-opus-5-5)
  2. 2026-10-17notes: Adopted its 2026-2027 roadmap (11 projects) on 12 Feb 2026. SGDSN says the Inter → SGDSN announced on 12 Feb 2026 that INESIA had adopted its 2026-2027 roadmap, wi — the sentence that INESIA has no separate legal personality or regulatory power was removed because no stored file supports it; the roadmap date is given as the date SGDSN announced the adoption2026-10-17-recheck-europe-2 · Claude Code (claude-opus-5-5)
  3. 2026-10-17summary: France's AI safety institute, created on 31 Jan 2025 before the Paris AI Action → France's national institute for AI evaluation and security, created on 31 Januar — rewritten from the mission page and the roadmap (NRBC rendered as CBRN, as SGDSN's English translation does): adds the roadmap's projects (tools for AI Act market surveillance authorities, synthetic-content detection, systemic risks from information manipulation, offensive cyber and NRBC proliferation, agentic systems) and says, as the page does, that SGDSN represents France in the AISI network2026-10-17-recheck-europe-2 · Claude Code (claude-opus-5-5)

Related records

EU AI Act European Union · Partially in force

Citation

Secrétariat général de la défense et de la sécurité nationale (SGDSN) and Direction générale des Entreprises (DGE). National Institute for the Evaluation and Security of Artificial Intelligence (INESIA). 2025-01-31. https://www.sgdsn.gouv.fr/nos-missions/anticiper-et-prevenir/evaluer-et-securiser-lia. Status: Published (non-binding, current); verified 2026-10-17 (level A). Local copy archive/FR-01/2026-10-17/official.html retrieved 2026-10-17 20:41:37 UTC, SHA-256 12e94c1a2b47e373fc104388860767418a9a73636d0c484ea7d08fa9676849fc. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record FR-01.

Improve this record

Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.