Does Replika train on your data?
Replika anonymises small portions of your messages and trains its own safety and chatbot models on them by default, with no way to refuse. Its outside model providers may not train.
No opt-out found in this policy
"This includes collecting and immediately anonymizing user feedback, and small portions of Messages and Content data to train our proprietary safety algorithms, enhance chatbot performance, prevent inappropriate outputs, and ensure compliance with safety standards."
- 6 Sept 2026
- 5 Oct 2026
- All accounts (free and Pro)
- global
- Fetched from the company
- CA13DD5E…66EA
How we read it: The training is on "small portions" of messages, "immediately" anonymised, for Replika's own safety and chatbot models. The same paragraph says the anonymised data "is not used to train third-party large language models or other AI systems", and the outside model providers that generate replies are bound "not [to] use your data for training their own AI models". None of that is a way out: the use rests on legitimate interests, the rights section names opt-outs for marketing, selling and sensitive data and a withdrawal of consent, and no setting or route touches it. The score sits at the band's floor for the anonymisation and the narrow purpose.
opt out, opt-out, object, settings, withdraw, choice
What the document does offer
- "Opt-out of marketing communications" and "Opt out of selling personal information and sharing for targeted advertising", as headed rights
- "Limit our use of sensitive personal information", a third headed right
- "Withdrawal of consent, where we have collected and processed your personal information with your consent" — the training use rests on legitimate interests, not consent
Why none of it applies: Four routes are named and none reaches the training use: the marketing, selling and sensitive-data opt-outs are about other processing, and the withdrawal of consent does not apply to a use the policy grounds in legitimate interests. No setting, form or address is named for refusing the anonymised training.
What each plan can refuse
The same questions, asked of every plan we could identify. Every cell comes from a quoted clause or a documented search - see the evidence above.
- None offered
- Not stated
- 84 / 100
- Yes, stated
- Unknown
How to opt out of Replika AI training
Plan by plan, from what Replika's own policy says about refusing. Where it publishes the steps, they're listed here with the limits they carry. Where there's nothing to switch off, or nothing graded, this says so instead of inventing a menu path.
All accounts (free and Pro)
No opt-out is stated in the policy on file for Replika.
What this policy has done since we started watching
The record starts here. Replika's policy was first captured on 6 Sept 2026, and there's no earlier copy to compare it against, so this verdict has no history behind it yet. From now on every re-read is diffed against that copy, and anything that moves appears here.
Everything we read
Vendor response
We wrote to Replika on 11 Sept 2026, quoting the clauses above and inviting a correction. Nothing has come back. That's published because a company declining to comment on its own policy is something a reader should know, and because the invitation should be on the record whether or not it's taken up. Any reply goes here, word for word.
Report a wrong clause