AI companions: the privacy checklist
These apps collect the most intimate data in consumer software. Here is what to check before you write anything personal.
By Editorial team
- privacy
- guide
An AI companion holds a category of data almost nothing else does: what you say when you think nobody is judging you. That deserves more attention than it usually gets.
Five things to check
- Training. Does the policy say your conversations may be used to improve models? Look for an opt-out available on the free tier, not just the paid one.
- Retention. How long are conversations kept after deletion? “Deleted immediately” and “removed from backups within 90 days” are not the same promise.
- Export. Can you download your history in a readable format? If not, it is not really yours.
- Third parties. Which analytics and advertising SDKs ship inside the mobile app? These apps often carry more than you would guess.
- Jurisdiction. Where is the company registered, and which regulator would you complain to? For a European user, an EU-registered entity makes GDPR rights effective in practice, not just in theory.
Good habits
Use a dedicated email address. Do not give a real full name, employer or address, even in roleplay. Assume everything you write could one day be read by a human: on most platforms, that is possible.
What a good policy looks like
The apps that score best on this criterion share three traits: a plain-language policy, an export button that works, and an explicit statement about training data. None of that is technically hard. Its absence is a choice.