Arthur Mensch, the founder and CEO of French AI company Mistral, is warning businesses to think twice before building on closed, proprietary AI models from rival labs.
According to The Decoder, Mensch argues that when companies feed their internal operations into a proprietary model, they hand the AI lab behind it a "front-row seat" to their business processes. His concern is that these labs are storing more and more customer data over time.
More pointedly, Mensch claims that some AI labs have, in certain cases, used that customer data to compete against those very same customers. In other words, the company running the model could turn around and use what it learns to enter your market.
The Decoder notes that the underlying concern is valid — questions about how much data closed AI providers retain, and what they do with it, are real. But the outlet also adds important context: Mistral has a clear commercial interest here, and it cannot really compete head-to-head with the leading frontier models. Warning customers away from closed rivals is also a way of steering them toward Mistral's own more open offerings.
The episode is a reminder that debates over "open" versus "closed" AI are not just technical or philosophical — they are also competitive marketing, with each side framing data and trust to its advantage.
Why it matters: As businesses rush to embed AI into their core operations, the question of who sees that data — and whether a supplier could become a competitor — is becoming a central part of choosing which model to trust.