Arthur Mensch, CEO and founder of the French AI company Mistral, has issued a stark warning to businesses regarding their reliance on proprietary, closed-source artificial intelligence models. In a recent LinkedIn post, Mensch argued that companies utilizing third-party closed models are effectively granting AI labs a “front-row seat” to their internal business processes and proprietary data.
According to Mensch, AI providers often store vast amounts of customer data, creating a risk where labs could leverage this information to compete directly with their own clients. “Frontier AI can accelerate the growth of your business, but if it’s not in your hands, it’s not going to be your growth,” he stated, advising companies to maintain control over their data, set independent access rules, and invest in training their own models.
Mensch’s perspective aligns with recent comments from Palantir CEO Alex Karp, who published a manifesto arguing that “controlling your weights is controlling your fate.” Karp contends that AI weights represent the distilled institutional knowledge of a company, and ceding control of these weights transfers the business’s competitive advantage—or “alpha”—to the AI provider.
While critics note that Mistral’s advocacy for open-source AI is tied to its own business model, recent industry experiments suggest there is merit to the argument. A joint project between Bridgewater and Thinking Machines Lab demonstrated that a fine-tuned open-source model (Qwen3-235B) could outperform frontier models in specialized financial analysis while operating at significantly lower costs. As the debate over AI sovereignty intensifies, the choice between convenience via proprietary models and control via open-source alternatives is becoming a critical strategic decision for enterprise leadership.