A debate to calm down
Artificial intelligence in France evokes a mix of fascination and mistrust. The organizers of the AI for All Week summarized it in three figures, taken from the Digital Barometer 2025 and the Pix Observatory of Digital Skills 2025: 48 % of citizens use AI, 56 % say they do not trust it, and 76 % have never been trained while feeling the need. Growing usage, low trust, and demand for training: this is the gap that the first edition of AI for All Week aimed to reduce.
The operation, held from 18 to 24 May 2026, was led by La Mednum, the cooperative of digital mediation actors, and Make.org Foundation, with the support of the Café IA mission, under the high patronage of Anne Le Hénanff, Minister Delegate for Artificial Intelligence and Digital Affairs. Its manifesto sets three missions: demystify AI, promote responsible use, and enhance digital mediation. The vocabulary is revealing: it is neither about selling AI nor imposing it, but about making it a subject of calm public discussion, close to citizens.
A decentralized mobilization
AI Cafés, introductory workshops, conferences, demonstrations, and citizen debates were held in media libraries, schools, universities, third places, post offices, and community halls. In Strasbourg, for example, the university's Digital Culture Center scheduled conferences, workshops, and meetings from 18 to 23 May. The Ministry of National Education was among the institutional partners, alongside the National Agency for Territorial Cohesion, France Travail, France Num, and the "Dare AI" plan.
The report published by the organizers in September 2026 reports 1 220 events, 645 organizing structures, and more than 39 000 citizens reached, with 87 % of events in person. The initial goal of 1 500 events and 50 000 people was not fully achieved, but the territorial anchoring is real, even in overseas territories, although the density of events remains uneven across regions.
Local authorities on the front line
One figure from the report deserves attention: 20,5 % of organizers were local authorities, alongside associations (27 %), companies (23 %), and training institutions. Departments, communities of communes, town halls, and municipal social action centers opened their doors to workshops.
This involvement extends a broader trend. According to the 2025 barometer of the Data Publica Observatory, published in November 2025, 49 % of local authorities with more than 3 500 inhabitants had already initiated an AI project, and 77 % will have done so by adding planned projects within the following twelve months. Generative AI represents 84 % of these projects. Far from being spectators, local authorities are experimenting with AI in internal management, user relations, or energy management.
The Greater Paris Metropolis provides an example. In April 2025, its council adopted a first strategy dedicated to AI and launched the "Metropolitan Appropriation" program, in three stages (understand and train, experiment, deploy), to support its 130 municipalities.
"I am delighted with the acceleration by the Greater Paris Metropolis in the field of Artificial Intelligence, a powerful driver of transformation for our municipalities." (Patrick Ollier, President of the Greater Paris Metropolis, press release of 7 April 2025)
Its vice-president for innovation and digital affairs, Geoffroy Boulard, then insisted on the need for a "regulated and controlled" AI and better "appropriation" by municipalities. The approach starts with training agents to identify use cases with them, rather than imposing tools from above.
The fear of AI as a "civil servant killer"
This approach addresses a documented concern. In a note published in December 2025, the National Union of General Directorates of Local Authorities (SNDGCT) noted that the fear of replacement could make AI perceived as a "civil servant killer." The same union estimated that 40 % of local authority agents were already using AI tools in an unregulated manner, a figure cited by the report of the general inspectorates on AI in administrations, published in July 2026. Moving away from this "shadow AI" requires providing secure tools and training agents, not just raising awareness.
Sovereign tool, an ambition to rebuild
The state has tried to offer its own solutions. The "Albert France services" experiment, led by DINUM and the National Agency for Territorial Cohesion, assisted advisors in 48 France services offices. In January 2026, DINUM announced that it would not be generalized "in its current form," and is now working on a tool adapted to the network, while its "AI Assistant," integrating Mistral models, was tested by 10 000 ministry agents. The episode reminds us that sovereign public AI remains a work in progress, and that local authorities, often alone facing commercial offerings, need shared solutions.
A Senate report that sets the stage
Local authorities, however, have benchmarks. The information report by Senators Pascale Gruny (Aisne) and Ghislaine Senée (Yvelines), adopted on 13 March 2025 by the Senate delegation for local authorities, is the first parliamentary report dedicated to AI in the local context. It recommends developing shared engineering between local authorities and with the state, raising awareness among elected officials and managers, and integrating ethical and environmental requirements into projects.
The limits of the exercise
We must keep perspective. A week of mediation, even successful, does not sustainably train the three-quarters of the population who express the need. The organizers acknowledge this: their projects for 2027 focus on earlier mobilization, better coverage of the least served territories, more practical and level-segmented workshops, and audiences still little reached, such as young people and job seekers.
The risk is also seeing the fracture that already separates territories reproduced at the local level: according to the general inspectorates, 90 % of metropolises are engaged in AI, but 73 % of municipalities with fewer than 3 500 inhabitants have no project.
A mediation that aligns with a European debate
The European framework itself has evolved. The AI Act required AI providers and deployers to ensure a sufficient level of AI literacy among their staff. Since the entry into force of the digital omnibus, on 27 July 2026, this obligation has become an obligation of means, and the Commission and member states must further promote and support this acculturation. Local mediation, led by associations and local authorities, thus becomes an essential relay for European policy.
The AI for All Week and the growing involvement of local authorities outline an original path: a democratic appropriation of AI, through mediation, training, and local experimentation. But this ambition will only fulfill its promises if it is sustained over time, has resources, and also reaches the most modest municipalities.
Sources: Report of the first edition of AI for All Week (September 2026); Manifesto of AI for All Week; Éduscol STI, April 2026; Digital Culture Center of the University of Strasbourg; Data Publica Observatory, barometer 2025; Banque des Territoires, presentation of the Data Publica barometer 2025; Greater Paris Metropolis, press release of 7 April 2025; Banque des Territoires, SNDGCT study, December 2025; IGF-IGAS-IGA report, July 2026; BFM TV / AFP, 9 January 2026; Senate, information report No. 447 (2024-2025); European Commission, AI literacy (Article 4).
