Tool page
Mistral AI
Mistral’s official consumer privacy policy (a French company) describes training on inputs and outputs, a user objection control, 30-day abuse-monitoring retention unless ZDR is on, and a separate controller/processor split for business use.
Partial page. The catalog URL mistral.ai/privacy/ 404’d; we used the official policy at legal.mistral.ai (effective September 3, 2026). That consumer privacy policy describes using Input and Output to train models and a user control to object to that use. It does not clearly label the default toggle state. The policy also says it does not apply when you use Mistral products to process personal data in a business context (Mistral as processor). API/ZDR exceptions are mentioned only as “except for specific APIs” and “unless zero data retention is activated,” without a full eligibility table in this policy.
Not a verdict. These are quotations from official vendor pages. They are not legal advice and not a determination that a plan is safe, compliant, or suitable for a particular kind of data.
Policy by plan
| Plan | Trains on inputs by default? | Opt-out / opt-in | Retention | ZDR | BAA / DPA | Data residency | Exceptions |
|---|---|---|---|---|---|---|---|
| Mistral AI Products for consumers (for example Vibe or Mistral AI Studio), where Mistral is the controller | Training is a stated purpose: “To train our artificial intelligence models (large language models) … according to context, Inputs … and Outputs.” A user control “allows you to object to the use of your input and output data for model training directly from your account.” The policy does not use the exact words “on by default” for that control. | Object to training on input and output data from your Mistral AI Account settings (the policy points to the Help Center for details). | “Except for specific APIs, we keep your Input and Output for the period necessary to generate the Output and then for thirty (30) rolling days to monitor abuse (unless zero data retention is activated).” Longer legal holds are also described. | Mentioned: “unless zero data retention is activated,” and “except for specific APIs.” Eligibility details are not spelled out in this privacy policy. | This policy “does not apply if you use our Mistral AI Products to process personal data in the context of your business activities. In this case, you are the data controller, and Mistral AI is the data processor.” Mistral’s help center points business customers to a DPA under the Data Processing Addendum of their agreement (not independently opened as a full legal text in this edition). | Not stated / not verified beyond Mistral being a company incorporated in Paris, France. | Third-party content displayed for features such as Web Search “is not used … to train our models.” Models are also trained on public internet data and third-party datasets, which is separate from your account inputs. |
Source quotes
“Except for specific APIs, we keep your Input and Output for the period necessary to generate the Output and then for thirty (30) rolling days to monitor abuse (unless zero data retention is activated). … We’ve introduced a user control which allows you to object to the use of your input and output data for model training directly from your account.” Mistral AI Products for consumers (for example Vibe or Mistral AI Studio), where Mistral is the controller · Checked 2026-11-18 · Mistral AI Privacy Policy (effective September 3, 2026)
Sources checked
- Mistral AI Privacy Policy
- Mistral terms hub
- Mistral docs
- Catalog: mistral.ai/privacy/ — 404 (URL)