August 18, 2026, (Inside AI) — France will hire only sovereign AI providers, such as Mistral, for government use, Budget Minister David Amiel said Tuesday. His statement explicitly excludes OpenAI.
“This excludes OpenAI,” he said.
The announcement follows a cyberattack on the French tax agency disclosed last Thursday. The Finance Ministry said data of about 700,000 taxpayers were stolen.
Amiel said earlier Tuesday that the government would use AI tools to identify vulnerabilities in its services to cyberattacks. The policy shift ties AI procurement directly to national security concerns.
Read: OpenAI and GSA Announce Free ChatGPT Access for All U.S. Government Levels
Tax agency chief Amelie Verdier said her services detected another data breach on Monday. That attack is still being evaluated, she added.
The decision marks a significant escalation in Europe’s push for digital sovereignty. France has long positioned itself as a leader in open-source and European AI alternatives.
Mistral, founded in 2023 by former researchers from Meta and Google DeepMind, has raised over $1 billion and is valued at roughly $6 billion. Its models compete with offerings from OpenAI and Anthropic.
By restricting government AI contracts to domestic providers, France aims to keep sensitive data within its jurisdiction. This approach mirrors similar moves in Germany and Italy, where public sector AI procurement favors local vendors.
The cyberattack context adds urgency. AI-powered threat detection could help agencies spot vulnerabilities faster, but only if the underlying infrastructure is trusted.
Critics argue that excluding global leaders like OpenAI could limit access to cutting-edge capabilities. However, French officials view sovereignty as non-negotiable for state systems.
The stolen taxpayer data reportedly includes names, addresses, and bank details. Such breaches have pushed European governments to reassess cloud and AI dependencies on U.S. companies.
Read: OpenAI Aligns Safety Practices with EU AI Act Implementation
Mistral’s flagship models, including Mistral Large and Codestral, are available via API and open weights. The company has emphasized compliance with EU data regulations.
France’s move could influence other EU member states. The bloc’s AI Act already imposes strict rules on high-risk systems used by public authorities.
No timeline was given for when the government would begin deploying sovereign AI tools. Amiel’s remarks suggest procurement guidelines will be updated soon.
The second data breach at the tax agency remains under investigation. Verdier did not specify which systems were affected or how many records were exposed.
Reporting by Inti Landauro; Editing by David Goodman, Alexandra Hudson.
