French AI and Digital Council report: protect without hindering, vulnerability in the age of generative AI
Protéger sans empêcher : nos vulnérabilités à l’ère de l’IA générative
The French Artificial Intelligence and Digital Council (CIANum) released the report of the "Generative AI and Vulnerability" committee, which reviews the vulnerability risks of generative AI and puts forward 18 recommendations; the committee was commissioned on February 20, 2026 by Anne Le Hénanff, France's Minister Delegate for AI and Digital Affairs, and held 55 hearings. The report says users develop a "cognitive debt" and emotional dependence on AI; according to MIT researchers, writing with chatbots reduces cognitive extent by 55%, and 92% of adolescents admit to using them longer than they intend; model bias, deepfakes and influence during election periods constitute societal-level risks, 16% of French people used generative AI to help them vote in the most recent municipal elections, and 73% of French people use it to obtain information. The committee recommends strictly constraining generative AI mechanisms that induce user dependence, making admission and retention conditions for mass-market generative AI systems subject to an ex ante vulnerability impact assessment plus ex post enhanced supervision, and carrying out universal AI education from early childhood.