Mistral's Data-Training Opt-Out: Consumers Must Toggle It Off Themselves

Mistral’s support documentation lays out how customers can stop their prompts, responses, and uploaded files from feeding the company’s model-training programs — and the defaults matter. Individual Vibe users are opted in unless they act: the switch lives under Admin panel → Vibe → Privacy, labeled “Allow your interactions to be used to train our models.” Enterprise Vibe customers get the opposite treatment, opted out by default with opt-in controlled at the admin level. On the iOS and Android apps, the equivalent control is a “Enable data sharing” checkbox under Data & Account Controls.

Anything attached or uploaded inside Vibe counts as input data, so documents flow into training unless the toggle is off — a detail worth noting for anyone pasting sensitive material into the assistant. A separate control governs Mistral Studio and the API: an “Anonymous improvement data” toggle in the Admin panel’s Privacy menu blocks API traffic from being used to improve the service.

The practical catch is that these controls don’t cascade. The Vibe and API opt-outs are independent, and disabling one leaves the other untouched, so administrators have to walk through and flip each switch deliberately. For organizations trying to keep proprietary data out of a vendor’s training pipeline, the burden is on them to audit every surface rather than trust a single global setting.