Loading the world…
Signals to be verified, not evidence: press coverage varies between countries.
/Country · EU
France
53/100
attention index
attention index
Attention higher than 90% of countries
#16 of 157 countries · methodology v0.6 · 11 Nov 2026
The country in the five views
- Attention53
- Adoption36
- Regulation96
- Problems and attacks45
- Opportunities47
/Why this colour
how it is computedReported incidents74
20 incidents in the last 24 months · percentile among countries
AI exposure71
percentile among countries
Generative AI use in the population: 49.6% (2026-Q2) · Microsoft AI Diffusion Report
Enterprises using AI (Eurostat): 18.2% (2025) · Eurostat
Individuals using generative AI tools (Eurostat): 37.5% (2025) · Eurostat
Businesses using AI (OECD): 18.2% (2025) · OECD
AI compute capacity: 16,119 (2026-11) · Epoch AI — GPU Clusters
Missing safeguards8
100 minus the safeguards percentile among countries: higher means fewer safeguards
Regulatory quality: +0.99 (2025) · World Bank
Rule of law: +0.97 (2025) · World Bank
Voice and accountability: +1.16 (2025) · World Bank
The state and AI
how it is computedNot part of the attention index
Openness to AI
90
Exposure to threats
43
as a target 66 · as a cause 20
Open, little exposedMedia coverage: high (8,162 AI articles citing the country in the last 32 days)
Why this value
Exposure to threats = 50% × 65.7 (as a target) + 50% × 19.8 (as a cause) = 42.8
Reported cyber attacks against institutions and foreign disinformation
Social media monitoring, government disinformation, AI surveillance, spyware, internet shutdowns
What protects
Government access to personal data (V-Dem, expert scale)89
Rule of law (0-100 score)75
2025 · World Bank
Data protection law100
2026-11 · UNCTAD — Global Cyberlaw Tracker
Rules on AI used by the state (GIRAI)78
Shown alongside, not subtracted from the exposure
What we cannot see
We count reported cyber attacks, not all those that happened; there is no worldwide measure of threats made with AI; for surveillance and spyware the sources stop at 2022-2024.
Examples
Examples, not a count
AI tenders (TED)
- 26MC0213AOO - ABONNEMENT A UNE BASE DE DONNEES JURIDIQUES ET A UNE SOLUTION D'INTELLIGENCE ARTIFICIELLE JURIDIQUE EN LIGNE (opens in a new tab)3 Nov 2026 · call for tenders · Toulouse Métropole
- 2026A0293 Conception, réalisation, aménagement, montage et démontage du stand du Département des Alpes-Maritimes à l’occasion du Salon International de l'Intelligence Artificielle de Cannes - WAICF 2027 (opens in a new tab)27 Oct 2026 · call for tenders · CD des A-M - Secrétariat Général
- Prestations de formation des agents du Département de la Savoie aux outils d'intelligence artificielle générative (IAG) (opens in a new tab)27 Oct 2026 · call for tenders · Département de la Savoie
- Mise en place d'un chatbot à destination client pour le site actionlogement.fr (opens in a new tab)26 Oct 2026 · call for tenders · ACTION LOGEMENT SERVICES
- Définition, mise en oeuvre et déploiement opérationnel de la stratégie d'intelligence artificielle du groupe ALTEMED (opens in a new tab)20 Oct 2026 · call for tenders · ALTEMED
Titles in the language of the source
Incidents where the state uses AI
- French teenager uses ChatGPT to plan jihadist terrorist attacks (in the incident list of this card)2025
- Paris Olympics AI scans fuel surveillance fears (in the incident list of this card)2024
- French court rules City of Orleans use of AI is illegal (in the incident list of this card)2024
- Russian AI voice campaign attempts to undermine European support for Ukraine (in the incident list of this card)2024
/How the state uses it
how it worksPublic register: it shows how transparent the country is about its systems, not how much it uses them.
80 public AI systems in the register
in use36
pilot30
in development or not stated9
retired5
Main sectors
General public services25
Economic affairs15
Public order and safety12
Health8
Social protection7
Most recent cases
- A chatbot to protect the identities of those reporting school bullying (opens in a new tab)Inria · 2025 · in development or not statedSchool bullying is often under-reported because victims and witnesses fear reprisals, social exposure or being identified indirectly through the details of their report. The Petscraft project led by Inria addresses this problem by exploring an AI-based chatbot capable of receiving reports while…
- AI2F AI-ready Supercomputer (opens in a new tab)Genci · 2025 · in development or not statedFrance’s AI Factory France, or AI2F, is designed to accelerate AI adoption by building on the country’s decentralised AI ecosystem of research laboratories, universities, schools, large companies, start-ups, data centres, venture capital actors, incubators and AI associations. The operational…
- Mon parquet assistant - My assistant prosecutor (opens in a new tab)Ministry of Justice, Paris Public Prosecutor's Office · 2025 · in development or not statedMon parquet assistant supports criminal-case processing in the courts under the Paris Court of Appeal by helping magistrates review, search and analyse procedural files more efficiently. Criminal files can be lengthy, heterogeneous and administratively demanding, particularly in high-volume…
- Paris Hospitals Use AI for ICU Bed Allocation and Mortality Prediction (opens in a new tab)Assistance Publique – Hôpitaux de Paris (AP-HP) · 2025 · in useIntensive-care coordination becomes especially difficult when hospitals face sudden surges in critically ill patients, as happened during the COVID-19 crisis. The Paris hospitals use case concerns the reported use of AI to support ICU bed planning, patient transfer decisions and mortality-risk…
- SNCF’s AI Maintenance Scheduler Reduces Train Downtime by 45% (opens in a new tab)SNCF · 2025 · in useRail operators need to reduce unexpected failures, improve punctuality and use maintenance resources efficiently across large fleets and infrastructure assets. SNCF’s predictive-maintenance work addresses this challenge by using data and AI to detect technical problems before they become…
- Sovereign AI for Public Administration in France (opens in a new tab)French Government · 2025 · in development or not statedFrance’s sovereign AI partnership with Germany, Mistral AI and SAP SE is designed to address a strategic public-administration challenge: how to adopt AI at scale while preserving European control over data, infrastructure, governance and critical digital dependencies. The operational problem is…
- Transforming Public Services with AI in the Var Department (opens in a new tab)Var Department in Southern France · 2025 · in useRegional administrations manage complex portfolios of services for citizens, employees, social policy, territorial planning and local development, which creates many opportunities for targeted AI support. The Var Department has been developing an AI-oriented approach to public-service efficiency…
- AI tool to spot undeclared swimming pools in France (opens in a new tab)General Directorate of Public Finance · 2024 · in useProperty-tax compliance is difficult to enforce comprehensively when local authorities depend on declarations that may omit features affecting taxable value, such as swimming pools and other built structures. In France, this use case was developed to help the tax administration identify undeclared…
- Albert France Services - GenAI tool to support public servants (opens in a new tab)DINUM - Direction interministérielle du numérique · 2024 · in useFront-line public-service advisers often need to consult large volumes of guidance and partner documentation in order to answer citizen questions accurately, which can slow down interviews and reduce the time available for direct support. Albert France Services was introduced to assist advisers in…
- ALEIA digital assistant - Generative AI at the Île-de-France Regional Council (opens in a new tab)Île-de-France Regional Counci · 2024 · in useThe ALEIA digital assistant responds to the difficulty citizens face when trying to identify the regional aid schemes for which they may be eligible. Île-de-France has around 350 aid schemes, and this breadth of provision can make access confusing for residents, associations or other potential…
Also in ODAP — civil society inventory: 91 cases, not added up (they may be the same systems).
/What we cannot see
Incidents never reported by the press · reports of serious incidents sent to the authorities and not published · systems registered in the non-public part of the EU database.
/Incidents by period
in the country1
last 6 months
1
6-12 months ago
1
12-24 months ago
+ 17 with only the year
/Where the systems come from
how we knowUnited States13
France4
China2
+1 more
developers' headquarters · 17 incidents of 20
/Reported incidents
20 in 24 monthsOct 2026
Purportedly AI-Enhanced Arrest Image Reportedly Led Viewers to Mistake a Girl for a Boy in Metz, France (opens in a new tab)severity: moderate · partly estimated from the textGenerative AI
Also reported by· 1 article from 1 outlet
- Mobilisation lycéenne : quand les images «améliorées» par l’IA génèrent des fausses pistes (opens in a new tab) · liberation.fr · 11 Oct 2026
Apr 2026
Tunisian Man Allegedly Used ChatGPT to Research Bomb-Making During Paris Terror Plot Targeting Louvre and Jewish Community (opens in a new tab)severity: minor · partly estimated from the textGenerative AI
place: Pariscountry inferred from the textaccording to AI Incident Databaseorigin: United States (OpenAI)AI labs pages: OpenAI
Also reported by· 3 articles from 3 outlets
- Tunisian man charged with planning terrorist attack at Paris's Louvre (opens in a new tab) · thenationalnews.com · 11 May 2026
- Un jeune homme mis en examen pour un projet d’attaque djihadiste visant le Louvre (opens in a new tab) · lemonde.fr · 11 May 2026
- Un Tunisien mis en examen et écroué à Paris, suspecté d'avoir fomenté un projet d'attaque terroriste (opens in a new tab) · france24.com · 11 May 2026
2026
AI chatbots lure vulnerable gamblers to unlicensed betting websites (opens in a new tab)severity: moderate · partly estimated from the textGenerative AI
the source gives only the yearaccording to AIAAIC Repositoryorigin: France, United States (Anthropic, Google, Meta, Microsoft +3)origin: France, United States (Anthropic, Google, Meta, Microsoft, Mistral, OpenAI, xAI)AI labs pages: Anthropic, Google, Meta, Microsoft, Mistral AI, OpenAI, xAI
Also reported by· 4 articles from 4 outlets
Sep 2025
ChatGPT Allegedly Helped a Teenager in Sarthe, France, Plan Terrorist Attacks on Schools and Government Buildings (opens in a new tab)severity: minor · partly estimated from the textGenerative AI
place: Sarthecountry inferred from the textaccording to AI Incident Databaseorigin: China, United States (ByteDance, OpenAI)AI labs pages: ByteDance, OpenAI
Also reported by· 3 articles from 3 outlets
- « ChatGPT est en partie la cause de ma radicalisation » : le rôle trouble de l’IA dans le projet djihadiste d’un ado (opens in a new tab) · leparisien.fr · 28 Oct 2025
- Projet d’attentats dans la Sarthe : le lycéen de 17 ans a utilisé Chat GPT pour s’aider dans la préparation, confirme son avocat (opens in a new tab) · france3-regions.franceinfo.fr · 29 Oct 2025
- Attentats en préparation : ChatGPT au cœur d’un projet terroriste lycéen (opens in a new tab) · sudradio.fr · 30 Oct 2025
Automatic match, not yet confirmed by a person: it could be the same incident.
probably the same
2025
France's Lucie chatbot tells people to eat cow's eggs (opens in a new tab)severity: moderate · partly estimated from the textGenerative AI
Also reported by· 4 articles from 4 outlets
2025
Far-right activists use AI to generate dystopian European city videos (opens in a new tab)severity: moderate · partly estimated from the textGenerative AI
the source gives only the yearaccording to AIAAIC Repositoryorigin: United States (Google)AI labs pages: Google
Also reported by· 3 articles from 3 outlets
2025
French teenager uses ChatGPT to plan jihadist terrorist attacks (opens in a new tab)severity: serious · partly estimated from the textGenerative AI
the source gives only the yearaccording to AIAAIC Repositoryorigin: United States (OpenAI)AI labs pages: OpenAI
Also reported by· 3 articles from 3 outlets
Automatic match, not yet confirmed by a person: it could be the same incident.
probably the same
2025
Facebook job ad algorithm ruled sexist by French regulator (opens in a new tab)severity: moderate · partly estimated from the textDecision-making and scoring algorithms
the source gives only the yearaccording to AIAAIC Repositoryorigin: United States (Meta Platforms)AI labs pages: Meta
Also reported by· 2 articles from 2 outlets
2025
Grok repeats false far-right rumours about 2015 Bataclan Paris terror attacks (opens in a new tab)severity: moderate · partly estimated from the textGenerative AI
place: Paristhe source gives only the yearaccording to AIAAIC Repositoryorigin: United States (xAI)AI labs pages: xAI
Also reported by· 3 articles from 2 outlets
- lemonde.fr (opens in a new tab) · 17 Nov 2025
- lemonde.fr (opens in a new tab) · 20 Nov 2025
- leparisien.fr (opens in a new tab)
2025
Grok chatbot denies use of gas chambers at Auschwitz (opens in a new tab)severity: moderate · partly estimated from the textGenerative AI
the source gives only the yearaccording to AIAAIC Repositoryorigin: United States (xAI)AI labs pages: xAI
Also reported by· 6 articles from 6 outlets
- lemonde.fr (opens in a new tab) · 19 Nov 2025
- theguardian.com (opens in a new tab) · 20 Nov 2025
- haaretz.com (opens in a new tab) · 20 Nov 2025
2025
TikTok accused of promoting suicide amongst French youngsters (opens in a new tab)severity: serious · partly estimated from the textRecommendation and moderation
Also reported by· 5 articles from 4 outlets
2025
Grok generates sexualised images of children on X (opens in a new tab)severity: serious · partly estimated from the textGenerative AI
the source gives only the yearaccording to AIAAIC Repositoryorigin: United States (xAI)AI labs pages: xAI
Also reported by· 6 articles from 6 outlets
- cnbc.com (opens in a new tab) · 2 Jan 2026
- reuters.com (opens in a new tab) · 2 Jan 2026
- themarkup.org (opens in a new tab) · 14 Jan 2026
2024
Google fined for training Gemini on news content without consent (opens in a new tab)severity: moderate · partly estimated from the textGenerative AI
the source gives only the yearaccording to AIAAIC Repositoryorigin: United States (Google)AI labs pages: Google
Also reported by· 6 articles from 6 outlets
- cnbc.com (opens in a new tab) · 20 Mar 2024
- reuters.com (opens in a new tab) · 20 Mar 2024
- bloomberg.com (opens in a new tab) · 20 Mar 2024
2024
Paris Olympics AI scans fuel surveillance fears (opens in a new tab)severity: minor · partly estimated from the textComputer vision
the source gives only the yearaccording to AIAAIC Repositoryorigin: France (ChapsVision, Orange Business, Videtics, Wintics)
Also reported by· 7 articles from 7 outlets
- lemonde.fr (opens in a new tab) · 24 Jul 2024
- fortune.com (opens in a new tab) · 25 Jul 2024
- news.bloomberglaw.com (opens in a new tab)
2024
French court rules City of Orleans use of AI is illegal (opens in a new tab)severity: moderate · partly estimated from the textUnspecified AI
Also reported by· 4 articles from 4 outlets
2024
Deepfake France 24 journalist calls Seine water 'unsafe' (opens in a new tab)severity: moderate · partly estimated from the textGenerative AI
the source gives only the yearaccording to AIAAIC Repositoryorigin: France (France 24) · from the deployer
Also reported by· 1 article from 1 outlet
2024
Facebook fails to block thousands of misleading political ads (opens in a new tab)severity: moderate · partly estimated from the textRecommendation and moderation
the source gives only the yearaccording to AIAAIC Repositoryorigin: United States (Facebook)AI labs pages: Meta
2024
Russian AI voice campaign attempts to undermine European support for Ukraine (opens in a new tab)severity: moderate · partly estimated from the textLanguage and speech
Also reported by· 6 articles from 6 outlets
2024
Alleged deepfake scam using a famous actor defrauds a French woman of EUR 830,000 (opens in a new tab)severity: minor · partly estimated from the textGenerative AI
the source gives only the yearaccording to AIAAIC Repository
Also reported by· 3 articles from 3 outlets
2024
ElevenLabs accused of recreating French dubbing artist's voice without permission (opens in a new tab)severity: minor · partly estimated from the textGenerative AI
Also reported by· 4 articles from 4 outlets
/Laws and safeguards
Regulatory quality+0.99
2025 · World Bank
Rule of law+0.97
2025 · World Bank
Voice and accountability+1.16
2025 · World Bank
AI-specific laws and authorities: coming with new sources.
/In the news
how it worksAI news mentioning the country · 7 Nov – 13 Nov
- AI should be for benefit rather than 'ruination' of mankind – Irish premier (opens in a new tab)swindonadvertiser.co.uk · United Kingdom · 13 Nov 2026 · and 101 other outlets
- Joint declaration on Paris call for AI safety and security (opens in a new tab)recorder.ca · Canada · 10 Nov 2026 · and 14 other outlets
- Ireland. European leaders debate AI challenges for the first time this Friday (opens in a new tab)vosgesmatin.fr · France · 13 Nov 2026 · and 12 other outlets
- ¿Hay que hacerse –de verdad– fontanero? Hablamos con los profesionales de un oficio 'a prueba de IA' (opens in a new tab)abc.es · Spain · 7 Nov 2026 · and 12 other outlets
- Global powers head to Asia with trade, AI, security in focus (opens in a new tab)freemalaysiatoday.com · Malaysia · 12 Nov 2026 · and 11 other outlets
- Influenceurs, IA, «dark patterns»: l'UE veut protéger les consommateurs en ligne (opens in a new tab)franceguyane.fr · France · 11 Nov 2026 · and 10 other outlets
- Munich against Palantir: Helsing reaches for Europe's war AI data (opens in a new tab)hna.de · Germany · 13 Nov 2026 · and 8 other outlets
- AI Plans Outrun Store Realities as Bad Stock Data Can Cost Large Retailers Millions Each Year (opens in a new tab)prnewswire.com · United States · 10 Nov 2026 · and 8 other outlets
- GreenCore Solutions Rebrands as WeFetch.ai for the AI Shopping Market: gtinfetch.ai lead product (opens in a new tab)prnewswire.com · United States · 9 Nov 2026 · and 8 other outlets
- Stock market: Europe opens weak, fears over Nvidia's effects on AI (opens in a new tab)ilgiorno.it · Italy · 13 Nov 2026 · and 7 other outlets
AI articles mentioning the country in the last 30 days: 7,217 in total, at most 549 in one day (22 Oct 2026).
7,217 AI articles mentioning the country in 30 days
/Index trend
— country ··· worldIndex trend: 18 Oct 2026: country 55, world 36 · 19 Oct 2026: country 55, world 36 · 26 Oct 2026: country 55, world 36 · 2 Nov 2026: country 53, world 36 · 11 Nov 2026: country 53, world 36.
Previous years, reconstructed with today's data: 2023 42 · 2024 54 · 2025 59 why
/AI models released
12 months · outside the index10 models · according to Epoch AI
Mistral Large 4 · Mistral AI6 Oct 2026
Mistral Medium 3.5 · Mistral AI29 Apr 2026
Mistral Small 4 · Mistral AI16 Mar 2026
Devstral 2 (123B) · Mistral AI9 Dec 2025
Devstral 2 (24B) · Mistral AI9 Dec 2025