Mistral's CEO has a warning: closed AI models get a front-row seat to your business
Arthur Mensch says companies relying on closed AI models hand over a window into their own operations — and a fresh finance experiment partly backs him up.
Arthur Mensch, founder and CEO of French AI lab Mistral, has some blunt advice for companies: if you run your business on someone else’s closed AI model, you’re giving that lab a very good view of how your business actually works. In a LinkedIn post this weekend, he argued that AI labs are storing more and more customer data — and that some “have a track record of going after their most successful customers thanks to this information.”
His recommendation: keep your data in open systems, set your own access rules, and — where it makes sense — train or fine-tune your own models, even if “these efforts might seem daunting.” His sharpest line: “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.” Palantir CEO Alex Karp made a similar case days earlier, publishing a manifesto arguing that “controlling your weights is controlling your fate” — weights being the trained model files themselves, the distilled result of everything a model has learned.
What’s behind this? Mensch is arguing his own book, and it’s worth saying so. Mistral is Europe’s only serious AI lab, it can’t match the raw performance of top models from OpenAI or Anthropic, and digital sovereignty — the idea that your data and tools stay under your own control — is exactly where Mistral stands to win. That said, a recent experiment gives his argument real weight. The hedge fund Bridgewater and Thinking Machines Lab fine-tuned Qwen3-235B, an open-weight model anyone can download, on their own internal investor evaluations. The result: 84.7 percent accuracy on financial document tasks, versus 78.2 percent for the best frontier model — at operating costs nearly 14 times lower. The catch: that comparison wasn’t independent, and big labs could likely close the gap by acquiring similar data.
What this means for you: If you’re just chatting with an AI assistant, not much changes — this debate is about companies wiring AI deep into their operations. But the underlying idea matters for everyone: the more of your work flows through a closed service, the more you depend on its provider’s goodwill and pricing. If you’re curious what the alternative feels like, running an open model locally with tools like Ollama or LM Studio is easier than it’s ever been. For businesses, the takeaway is more concrete: your own expert knowledge, used to tune an open model, can beat the giants in your niche — worth knowing before you sign the next big AI contract.
Sources
Hollywood wants ByteDance's AI video tool banned — and quietly keeps using it
The LA Times reports studios are fighting Seedance with cease-and-desist letters while tolerating its use on a 'don't ask, don't tell' basis — a snapshot of an industry split on AI video.
