Generative AI, a major challenge for local communities
Faced with the rapid growth of generative artificial intelligence (AGI), many public stakeholders are seeking a balance between innovation, practical utility, and responsible risk management. Through the Rennes experience, it is possible to grasp the profound transformation that AGI imposes on local authorities, both in its promises and its adaptation requirements, but also in the multiple challenges it raises.
Generative artificial intelligence has crossed a threshold of integration into everyday practices, both private and professional. Public officials, like citizens, are exposed to it daily. Rennes, far from giving in to a passing fad or simple tech-savvy adoption, is promoting a proactive and structured approach: it's not just about appropriating technical tools, but also about preserving the public interest, transparency, equal access, and data sovereignty. The methodology deployed is based on a systemic and multidisciplinary analysis, bringing together experts in digital, human resources, legal, and communications, in order to draw up a comprehensive strategic diagnosis. The objective? To foster a shared culture of risk and opportunity management and equip all public officials for the informed use of AGI.
One of the report's first findings is the need for ongoing monitoring and constant adaptation: AI is evolving at an accelerated pace, both in terms of uses and the technologies themselves. In this context, the Rennes community has chosen to replace a one-off mission with a permanent observatory, responsible for capitalizing on and disseminating best practices, ensuring regulatory monitoring, and serving as an entry point for all questions regarding AI usage or projects.
From a political and economic perspective, AI is now a matter of national sovereignty and massive investment. The world's major powers are placing AI at the heart of their power strategies, through the creation of consortiums, the establishment of dedicated infrastructures, and the launch of large-scale projects. But for local authorities, this international context implies pragmatic choices: which solutions should be adopted (American, European, open source, specialized, etc.)? What guarantees should be expected regarding confidentiality, GDPR compliance, cost control, or hosting sovereignty? The report highlights the need for a thorough evaluation of each offer, emphasizing the advantages of sovereign or localized solutions, as well as the need to maintain internal audit and management capabilities.
One of the major axes of the approach lies in risk management. The risks identified are multiple and affect both the environment (exponential energy impact of data centers, greenhouse gas emissions, overconsumption of resources), data (potential leaks, poor management of personal data, loss of control over the flow of information), practices (uncontrolled use or misuse by third parties, atrophy of skills, stress linked to the increasing complexity of tools) and the very economy of communities (risks of dependence on licenses, hidden costs, changing economic models). Faced with these challenges, risk mitigation rests on several pillars: awareness and acculturation of agents, limitation of unjustified uses, adoption of frugal tools, rigorous contractualization with service providers, sovereign hosting, documentation and systematic traceability of uses.
The regulatory aspect is equally crucial. The entry into force of the European regulation on artificial intelligence (AI Act) imposes a classification and framework for AI based on its risk level, with differentiated obligations. This new framework complements that of the GDPR, intellectual property law, charters for responsible digital use, and the law for a digital republic. For local authorities, achieving compliance must therefore rely on strengthened legal oversight and the appointment of specialized contacts (Data Protection Officers, AI liaisons) who guarantee respect for users' fundamental rights and the traceability of automated decisions.
In terms of uses, AGI is divided into two broad categories: generic, cross-functional uses, which relate to routine administrative tasks (drafting and summarizing documents, translation, knowledge management) and specialized uses, closely linked to professions (human resources, accounting, technical project management). The challenge for the community is to clearly distinguish between these two families, to experiment in safe and controlled spaces, and to systematize the evaluation of the real impact on the activity of agents. This is not only to avoid any massive substitution of humans by machines, but also to promote AGI as an assistant, a decision-making aid, and not as a factor of dehumanization or professional downgrading.
A particular effort is being made on acculturation and continuing education: digital literacy is becoming a basic skill, to be developed around the ability to use tools, to question their results, to exercise critical thinking, and to understand and respect the regulatory framework. This involves workshops, training courses, the creation of Open Badges, and the integration of IAG into the integration systems for new agents.
Technologically, the range of solutions available is vast and constantly evolving. American solutions (ChatGPT, Gemini, Claude, etc.) offer power and functionality, but raise questions about dependency, cost, and sovereignty. European solutions (Mistral, DelibIA, government solutions) put forward arguments for compliance and regulatory proximity, but remain to be consolidated in terms of functionality, robustness, and user community. Installing proprietary models, on local infrastructure or in the cloud, offers the dual advantage of control and customization, but requires significant resources in management, maintenance, technical expertise, and user support. The recommended approach tends toward the adoption of hybrid, scalable solutions capable of ensuring alignment with responsible digital values and data security.
The originality of the Rennes approach also lies in the establishment of an observatory, the true backbone of IAG management. This observatory, composed of a multidisciplinary steering committee and technical groups, is responsible for ensuring technological monitoring, capitalizing on feedback, leading the user agent community, supporting skills development, updating charters and monitoring quantitative and qualitative indicators (number of projects, adoption rate, satisfaction level, efficiency gains, maturity of services).
This overview invites each public actor to ask the right questions and to go beyond the simple experimentation stage: how can we guarantee real sovereignty over infrastructures and data? How can we train and support all agents, so that they remain actors and not spectators of the digital transformation? How can we make digital technology a lever for innovation and efficiency without ever sacrificing the human dimension or democratic requirements? So many avenues to follow, so that AGI is fully a common good, controlled and shared.
To learn more: discover our overview of AI in communities and territories and our analysis of France faces the challenge of generative AI.