Glonce

Mistral explains how to opt out of Vibe and API data training

securityThe Glonce editorsHacker News🔥59Glonce rating59Impact40Novelty25Relevance75Surprise15Players55Credibility88

Mistral published a help center article explaining when a user's input and output data, meaning conversations, documents, and other user-provided content, may be used in its model training programs, and how to opt out. The rules differ by product. In Vibe, individual users are not opted out by default: their interactions can be used for training unless they turn that off themselves, which they can do at any time in their own settings. Vibe Enterprise customers work the opposite way: they are opted out of training by default, and the opt-in toggle is controlled at the admin level rather than by each employee. On desktop, opting out means going to the Admin panel, opening the Manage section, selecting Vibe, then in the Privacy section disabling the toggle labeled "Allow your interactions to be used to train our models." The same setting exists on iOS and Android: open the app's Settings page, go to Data & Account Controls under Account, and deselect the "Enable data sharing" checkbox. The article notes that documents attached or uploaded within Vibe count as input data too, so a user who wants those documents excluded from training needs to opt out the same way. Once an opt-out is confirmed, Mistral says it stops using that user's input or output data for training purposes. A separate set of instructions covers Mistral Studio and the API: customers there also retain full control and can opt out at any time, through the Admin panel's Privacy menu, under a section called "Anonymous improvement data," by disabling the toggle that allows API calls and related data to be used to improve Mistral's services. Mistral is explicit that the Vibe and API opt-out toggles are separate: turning one off does not turn the other off, so each has to be configured individually.

Key facts

  • Vibe users are opted into training by default and must actively turn it off in their own settings; Vibe Enterprise customers are opted out by default, with the toggle controlled at the admin level instead.
  • Desktop opt-out for Vibe: Admin panel, Manage, Vibe, Privacy, then disable "Allow your interactions to be used to train our models."
  • Mobile opt-out for Vibe (iOS and Android): app Settings, Data & Account Controls under Account, then deselect the "Enable data sharing" checkbox.
  • Documents attached or uploaded within Vibe count as input data and fall under the same opt-out.
  • Mistral Studio and API customers opt out separately, via the Admin panel's Privacy menu under "Anonymous improvement data"; the Vibe and API toggles are independent and each must be set on its own.

Why it matters

AI products routinely use the data that flows through chat interfaces, coding tools, and APIs to improve their models, and users are often unclear on what actually happens to what they type or upload. Mistral's help article makes the mechanics explicit for its two main surfaces, Vibe and the Studio and API stack, stating plainly which category of user is opted in by default and which is opted out, and exactly what to click to change it.

Who it affects

Anyone using Mistral's Vibe product, whether as an individual consumer account or as an employee under a Vibe Enterprise deployment, and customers of Mistral Studio or the Mistral API who want to control whether their calls feed model training.

How to use it

For Vibe on desktop, go to the Admin panel, open Manage, select Vibe, and in the Privacy section disable the toggle labeled "Allow your interactions to be used to train our models." On iOS or Android, open the app's Settings page, go to Data & Account Controls under Account, and deselect the "Enable data sharing" checkbox. For Mistral Studio and the API, the control sits in the Admin panel's Privacy menu under "Anonymous improvement data"; disabling that toggle stops API calls and related data from being used to improve Mistral's services. The two toggles are independent, so opting out of Vibe does not opt a user out of API-based training, and the other way around.

How solid is it

This is Mistral's own help center documentation, so the settings it describes and where to find them are first-party and about as authoritative as this kind of information gets. What is missing is context around it: the article carries no publication or last-updated date, and it does not name a spokesperson or executive behind the guidance.

Risks and caveats

The article does not say whether plain, non-Enterprise Mistral API or Mistral Studio accounts are opted in or opted out of training by default, only that Vibe consumer accounts are opted in and Vibe Enterprise accounts are opted out. It gives no retention or deletion timeline for data already collected before a user opts out, does not say which models the data trains, and says nothing about whether opting out affects service quality, pricing, or feature availability.

“Once the opt-out is confirmed, Mistral no longer uses your input or output data for the purpose of training its models.”

— Mistral help center