AI Infrastructure Market: Mistral CEO Warns Against Closed AI Model Vendor Lock-In

Arthur Mensch, co-founder and CEO of Mistral AI, published a LinkedIn post urging enterprise leaders to avoid dependence on proprietary, closed AI models. He argues that providers of such models gain "immense leverage" by observing how companies integrate AI into their business context, learn from that data, and may eventually compete with their most successful customers. Mensch recommends building independent AI infrastructure using open-source models, storing data in open systems, enforcing strict AI access management, and creating a continuous internal training flywheel. He also promotes Mistral's "Studio" control plane and "Forge" training platform, emphasizing sovereign, zero-data-retention deployments. The article notes that his claims partly overlap with documented cases, such as Anthropic cutting off Windsurf's model access, but says he offers no evidence for deliberate customer targeting. Palantir CEO Alex Karp has also recently warned against vendor lock-in in enterprise AI.