Mistral AI Now Trains on User Prompts by Default

Mistral AI has flipped its default privacy setting. Free and paid users now have their input and output data used for model training unless they manually opt out. Enterprise tier is exempt. That's a real change from the previous opt-in stance, and it affects anyone on Mistral's chat or API plans.

What changed in Mistral's data policy?

Mistral's official help article now says non-enterprise user data is used for training by default. Users can switch it off in their account settings, but the default is now sharing. The article doesn't say when the change took effect or how many users were affected, but the policy is live.

What does this mean for developers?

If you're on Mistral free tier or a paid non-enterprise plan, assume your prompts could end up in the training set. Enterprise customers are excluded. Developers handling sensitive code or personal data should review their settings and opt out where needed. For teams with strict compliance requirements, this is a good moment to check your data-processing agreements.

This looks like a deliberate move to grow training data at a time when high-quality human feedback is scarce. The lack of fanfare makes it easy to miss — but for Mistral users, the fine print just got a lot more important.