The French government wants to use artificial intelligence to identify vulnerabilities in administrative systems, whilst stipulating that it will rely on ‘sovereign’ suppliers, such as Mistral AI. Minister David Amiel stated explicitly on 18 August that this means OpenAI will be excluded from this area of procurement.
The statement came following a cyberattack on the French tax administration, the DGFiP. According to the Ministry of Finance, data relating to 678,000 individuals and organisations was stolen, including tax and land registry information. A further breach of the systems was detected on 17 August.
However, this is not a sudden U-turn. Back in June, France launched the ‘Notre IA’ plan, and the administrative tool L’Assistant, developed by DINUM using Mistral AI technology and the SecNumCloud infrastructure, was made available to civil servants. The strategy aims to boost productivity, but also to reduce dependence on foreign technology platforms.
The significance of this decision therefore extends beyond cybersecurity. France is increasingly using public procurement as an element of its industrial policy. Government contracts can provide European AI companies with revenue, track records and the scale needed to compete with US suppliers. This approach is in line with the EU’s technological sovereignty package of June 2026, which also includes mechanisms for assessing the sovereignty of cloud services and AI.
However, favouring local suppliers may restrict competition and access to the most advanced models. Nor will the mere choice of European AI solve the problem of outdated systems or weak access controls. The French case illustrates a broader shift: AI in public administration is becoming not only a tool for efficiency, but also a component of national security and economic policy.
