Arthur Mensch says enterprises need open models, open data, and their own training flywheel, an argument that conveniently ends at Mistral’s products
Mistral CEO Arthur Mensch used a LinkedIn post to argue enterprises must adopt open-source models, open data systems, and their own training flywheels, warning closed providers gain “immense leverage” over customers. His data retention and customer-competition claims have real but caveated anchors, and the argument doubles as a pitch for Mistral’s Studio and Forge products.
Arthur Mensch, cofounder and chief executive of French AI lab Mistral, has urged enterprise leaders to abandon closed AI models. In a LinkedIn post, he argued that closed providers are now forcing data retention and gaining “immense leverage” over their customers’ businesses.
As companies connect models to their internal context, Mensch wrote, providers see it, learn from it, and have a history of going after their most successful customers. The sharpest part of that charge, that providers use customer information to pick their targets, is an inference he offered no evidence for.
The data retention claim has a real anchor, with caveats. A US court ordered OpenAI to preserve ChatGPT logs during The New York Times copyright case, though enterprise and zero-data-retention API customers were excluded and the blanket order was later lifted.
The customer-competition worry is better documented. Anthropic cut off coding startup Windsurf’s model access in 2025 while building its rival Claude Code, and Brookings has warned that model providers increasingly compete with their own customers as they chase application-layer revenue.
Mensch’s programme runs from open models to open data stores, strict access controls, and a continuous training flywheel that improves systems on internal interactions. The goal, he wrote, is turning the edges of a business into AI systems that vendors and competitors cannot replicate.
He was candid that this amounts to a complete replatforming of IT and a change in how companies operate. Access control is a particular minefield, he argued, because AI models excel at surfacing information that employees were never meant to see.
Training your own models is no longer a fringe position. British startup Cosine has rallied BT, HSBC, and BAE Systems to build a sovereign UK frontier model, while Palantir has published an AI sovereignty manifesto taking aim at the big labs.
Mensch’s argument lands, not coincidentally, on Mistral’s own products. The company sells Studio, a control plane for building and governing AI systems, and Forge, a custom model training platform it launched in March.
Mistral deploys on customers’ infrastructure or through hosted services it says retain no data. The pitch targets European enterprises already anxious about dependency on US providers, an anxiety that has powered the continent’s sovereignty push and Mistral’s rise with it.
The Paris-based lab is reportedly in funding talks at a €20bn valuation and recently launched an industrial AI stack with Airbus, BMW, and EDF as launch customers. It profits directly if enterprises accept the argument, which does not make the argument wrong.
Mensch closed by warning that frontier AI only accelerates your growth if it is in your hands. For Europe’s biggest open-weights lab, hands and business model happen to point the same way.
Get the most important tech news in your inbox each week.
