European companies turn to private AI to protect sensitive data

European companies including Airbus, TotalEnergies and BMW are adopting privately controlled AI systems to keep sensitive business information away from external providers. Concerns about confidentiality, model training and changing service terms are encouraging businesses to run models in private clouds or on local servers, sometimes disconnected from the internet.

Mistral is benefiting from this demand for sovereign AI. ASML, TotalEnergies and Airbus have integrated proprietary data into its models, while BMW has worked with the French company on a model that accelerates vehicle crash-test evaluation. Mistral has also partnered with Cloudera, whose data-management platform lets businesses use models from different providers within their own security boundaries.

Open-weight models, which companies can download and adapt, offer an alternative to sending information to external cloud services. European regulation and efforts to reduce dependence on foreign technology are reinforcing that approach, particularly in critical industries. However, fully European AI remains out of reach because of dependencies including Nvidia chips, and businesses have limited domestic model choices.

A Cloudera report based on surveys of more than 1,500 technology leaders worldwide found that nearly half of businesses have already moved part of their AI operations to private systems. The shift reflects a broader priority: retaining control over where corporate data is stored, who processes it and which systems can access it.