ABSTRACT
Authors: Émilie AGNOUX and Johan THEURET, Le Sens du service public for the Jean Jaurès Foundation
Starting from the observation that artificial intelligence is already present in public services and government departments, the authors examine the impacts of this rapid adoption and, in particular, whether it will remain compatible with the principles of public service.
The use and adoption of AI in public services are advancing faster than the regulations intended to govern them (nearly 40 per cent of public sector staff are reported to be using ‘shadow AI’). The arrival of ‘agent-based’ AI – systems capable of breaking down a task, using several tools simultaneously and acting with greater autonomy – also raises the question of accountability, as the line between assistance and decision-making becomes blurred. The authors identify two risks for professional roles and the relationship with users: excessive trust in AI or its rejection due to a lack of understanding. They also highlight our dependence on international players and cloud infrastructure, raising issues of sovereignty, cost control and environmental impacts. The authors consider that the European framework (AI Act – 2024) is useful but insufficient, as it sets out legal obligations without defining a political vision for the role of AI. Similarly, the ‘Notre IA’ plan remains focused on announcements and use cases, without fully defining its objectives or assessing its long-term effects.
The authors propose using technology to enhance human capabilities rather than merely automating tasks and reducing human intervention (enhancement versus automation). Four principles of governance are put forward: transparency, human accountability, sovereignty and compulsory training. They also call for projects to be assessed on the quality of the service provided, not merely on productivity.
Keywords: administration, AI agents, automation, augmentation, local authorities, Cloud Act, digitisation, staff training, governance, AI Act, agent-based AI, generative AI, artificial intelligence, environmental impact, innovation, innovations, tools, public policy, the ‘Territoires d’IA’ programme, accountability, public services, shadow AI, digital frugality, AI solutions for the public sector, digital sovereignty.
