Mistral documents how to opt out of training on your inputs and outputs

Mistral's help centre carries an article explaining that in certain cases customer input and output data — conversations, documents and other user-provided content — may be included in the company's model training programmes, and that users can opt out at any time. The procedure differs by service.

Vibe, Teams and Enterprise

For Vibe, users are not opted out by default and can change that whenever they choose; in Vibe for Teams, administrators can disable data training for the whole organisation, and in Vibe Enterprise customers are opted out by default with the toggle managed at admin level. In the admin panel at admin.mistral.ai, a user selects Vibe under Manage and, in the Privacy section, disables the toggle labelled “Allow your interactions to be used to train our models”. Mistral notes that documents attached or uploaded inside Vibe count as input data, so opting out also covers those files.

Mobile apps

On iOS and Android the path is shorter: open the Settings page, select Data & Account Controls under the Account section, and deselect the “Enable data sharing” checkbox.

Studio and the API

Mistral Studio and the related API services have their own control. In the admin panel the customer opens the Privacy menu in the left-hand navigation and, under “Anonymous improvement data”, disables the toggle to stop API calls and related data from being used to improve Mistral's services.

The documentation stresses that the Vibe and API toggles are separate: opting out of one does not opt out of the other, and each has to be configured individually. Once an opt-out is confirmed, Mistral states that it no longer uses the customer's input or output data to train its models.