Exposed: Mistral AI

Mistral AI is a French AI company that has rapidly emerged as a leading developer of open-weight large language models, positioning itself as a European alternative to US-dominated AI development. Founded by former Google DeepMind and Meta AI researchers, Mistral raised over $2 billion within its first 18 months. The company releases model weights openly (Mistral 7B, Mixtral, Mistral Large), enabling local deployment that avoids cloud data collection. However, Mistral also operates La Plateforme, a commercial API service, and Le Chat, a consumer chatbot that collects conversation data on Mistral servers. Like all major AI companies, Mistral trained its models on web-scraped data without universal consent. Mistral has benefited from EU regulators who championed it as a homegrown AI champion, leading to concerns that the company received favorable treatment during EU AI Act negotiations. A leaked early version of Mistral Large was distributed without safety guardrails, demonstrating the dual-edged nature of open-weight releases. Despite its European base and GDPR proximity, Mistral faces the same fundamental tensions between commercial AI development and privacy that affect all players in this space.