Privacy policy

Protecting your privacy is at the heart of Ozer. This policy describes, in accordance with the General Data Protection Regulation (GDPR) and the French Data Protection Act, the data processing carried out as part of the Service.

1. Data controller

REALITIM, 34 rue Joncours, 44100 Nantes, France. For any question about your data: dpo@realitim.com.

2. Data collected

Some data relates to sexual life or orientation and constitutes sensitive data within the meaning of article 9 of the GDPR. It is only processed on the basis of your explicit consent, collected at sign-up and reminded before any data is sent to an AI provider, and can be withdrawn at any time.

3. Purposes and legal bases

4. Recipients and processors

Your data is never sold. It is shared only with technical processors acting on our instructions:

• Google Firebase / Google Cloud (authentication, database, hosting, Google Analytics, Cloud Vision and Video Intelligence for media moderation, Vertex AI Gemini for audio/video understanding, and Vertex AI for opt-in semantic memory);
• Anthropic (Luna replies, quiz analysis, recaps, challenge ideas and safety review — commercial API: your exchanges are not used to train their models; they may still be processed temporarily for security, abuse prevention and legal compliance);
• NanoBananaAPI (nanobananaapi.ai — image generation from your written prompt and, in retouching mode, the photo you provide);
• Black Forest Labs (image generation, processed in the European Union — receives the reference photo when you request an image based on one of your photos);
• Meta (Facebook / Instagram campaign measurement, including via RevenueCat, only with your in-app consent);
• RevenueCat / Apple / Google (subscription management and, with your consent, advertising attribution);
• Sentry (stability logs);
• Apple and Google (system dictation in the composer; notifications via APNs and Firebase Cloud Messaging);
• your partner, for the sole data you choose to share in couple mode.

In the app, a separate screen asks you to accept or decline usage measurement and attribution (GA4, Meta, RevenueCat); you can change this in Profile. On iOS, Apple separately shows App Tracking Transparency before any IDFA is collected. Declining ATT does not block an analytics consent without an identifier. Without tracking, attribution may rely on aggregated mechanisms (e.g. SKAdNetwork).

Before you decide. Between installation and the moment you answer that screen, measurement runs in a degraded mode (Google’s “advanced consent mode”): no advertising identifier, no persistent identifier written to your device, and nothing sent to Meta. Only aggregated measurements, not linked to any individual, are sent, so that we can tell how many installs come from a campaign. If you accept, measurement becomes complete; if you decline, it is switched off entirely, including that degraded mode.

4a. Artificial intelligence processing

Your consent before anything is sent. Before your data is transmitted to an AI provider, the app tells you what data is sent, to whom and why, then asks for your explicit consent (“Allow” or “Don’t allow”). The notice appears on first launch after sign-in, then before each AI feature until you have agreed. Without your consent, no data is sent to AI features (Luna, analyses, recaps, ideas, image generation and retouching). The date and version of your consent are stored with your account.

Data concerned: your messages and answers (quizzes, challenges, daily question, recaps), the photos, voice notes and videos shared with Luna, your image prompts and reference photos, as well as elements of your profile and couple (first name, preferences, Luna’s memories) to personalise replies.

Recipients: Anthropic, Google Cloud (Vertex AI Gemini, Cloud Vision, Video Intelligence), NanoBananaAPI and Black Forest Labs, depending on the feature used. We only use providers whose commercial terms offer protection equivalent to ours and rule out training their models on your data.

Several AI providers, used only server-side — not to train their models, under the applicable commercial terms. No API key is stored on your device.

Anthropic (commercial API) receives your messages and the excerpts needed for Luna’s replies, quiz analysis, recaps, challenge ideas and a safety review of Luna’s replies.

Google Cloud — Vertex AI Gemini transcribes voice notes and describes videos so Luna can understand them. Cloud Vision and Video Intelligence automatically moderate photos and videos. If semantic memory is enabled, Vertex AI turns text excerpts into embeddings, without training third-party models.

The app also offers AI image generation (photo editor, challenge proofs, story illustrations), handled by NanoBananaAPI (Nano Banana 2 model). Your written prompt and, in the retouching modes, the photo you provide are sent from our servers through a temporary link, for as long as it takes to produce the image. Generated images carry an “IA” badge burned into the image and can be reported directly from the app, even before being shared. Prompts aiming to produce images of minors, of real people without their consent, or sexually explicit content are blocked before anything is sent to the model.

Dictation in the composer uses the system speech recogniser (Apple or Google), preferably on-device; otherwise the system dictation service may process the audio just long enough to transcribe it. Generated text and images (including quiz analyses, recaps and challenge ideas) can be reported in the app; those reports are used to refine filtering.

Automatic media moderation (Cloud Vision, Video Intelligence), which is necessary for everyone’s safety, also applies to media exchanged with your partner.

Withdrawing your consent. You can withdraw it at any time in the app: Settings → Luna → AI processing. Sending stops immediately: Luna stops replying and AI features close. What has already been processed is not erased by the withdrawal: use “Clear my history” and Luna’s memory, or delete your account.

5. Transfers outside the European Union

Some processors (Anthropic, NanoBananaAPI, Meta and certain Google services) are located outside the European Union, mainly in the United States. These transfers rely on the safeguards provided by each provider’s terms (standard contractual clauses of the European Commission and/or the Data Privacy Framework).

6. Retention periods

• Account data, messages, media and generated content: for as long as the account exists, then purged within 30 days of deletion;
• Reports (reason, frozen copy, and the media if any): for the time needed to review them and meet legal duties, then deleted or anonymised;
• Analytics and attribution events, if you consented: the retention of the providers concerned (Google, Meta, RevenueCat);
• Billing data: kept in accordance with legal accounting obligations;
• Sentry logs: for diagnostics, not tied to your conversations.

7. Your rights

You have the rights of access, rectification, erasure, restriction, objection and portability, as well as the right to withdraw your consent at any time. See our Your rights (GDPR) page. You can exercise these rights at dpo@realitim.com or delete your account directly from the app.

In particular, you can withdraw at any time your consent to your data being sent to AI providers (Settings → Luna → AI processing).

You can lodge a complaint with the CNIL (www.cnil.fr) or with your local supervisory authority.

8. Security

We implement appropriate technical and organizational measures (encryption, strict access control, minimization) to protect your data. AI processing runs server-side; no sensitive key is stored on your device.

9. Cookies

The site uses trackers described in our cookie policy.

10. Contact

For any question about your data, to exercise your rights or report an incident: dpo@realitim.com.