Mistral changes default: user data used for training unless opted out
Original: Mistral now trains on user input by default, except on enterprise tier
Why This Matters
Default-on training data collection affects millions of standard users and raises data privacy considerations for businesses evaluating Mistral's consumer tiers.
Mistral has updated its data governance policy, enabling model training on user input and output by default for standard and Pro users of its Vibe platform. Enterprise customers remain opted out by default, with admin-level controls available.
Mistral has updated its help documentation to clarify that user input and output data — including conversations and uploaded documents — may be used in its model training programs by default. This applies to standard and Pro-tier users of its Vibe platform (formerly Le Chat). Users can opt out at any time through their personal settings.
Enterprise customers are treated differently: they are opted out of training by default, and any opt-in must be toggled at the admin level via the Admin panel under 'Manage > Vibe > Privacy.' The setting in question is labeled 'Allow your interactions to be used to train our models.'
Mistral also explicitly notes that documents attached or uploaded within Vibe are classified as input data and are subject to the same training use policy, urging users who handle sensitive files to opt out. Once an opt-out is confirmed, Mistral states it will no longer use that user's data for training purposes.
The policy change represents a shift from a previously more opt-in-oriented approach, moving the default toward data collection for standard users while preserving enterprise protections.