Privacy Policy

Last updated: July 27, 2026

Tavi (“Tavi: AI Reply Assistant”) is an iOS app developed by OzTech App Studio (Istanbul, Türkiye). This policy explains exactly what data the app handles, who processes it, why, and the control you have. The short version: Tavi has no sign-up and never asks for your name, email or phone number. A screenshot leaves your phone only after you have said yes, it is sent to Anthropic with nothing attached that identifies you, and it is never stored — not by us, not in our database, not in our file storage.

1. Who is responsible

The data controller for Tavi is OzTech App Studio, an independent app studio based in Istanbul, Türkiye. The app is distributed through the Apple App Store. For any privacy question or request, contact support@oztechapps.net.

2. The consent gate — read this one first

Tavi's core feature sends a screenshot of a conversation to an AI service. Because that screenshot may contain another person's words, it is placed behind an explicit gate rather than buried in this page:

• You are asked before anything is sent. The first time you try to analyse a screenshot, Tavi shows a consent screen that says what will be sent, that it goes to Anthropic by name, and what happens to it. Nothing is uploaded until you agree.
• Declining leaves the app usable. If you say no, no upload starts. You can browse the app, change settings, and come back to it later.
• The gate is enforced on the server, not just in the screen. Every analysis re-reads your consent flag before the image is passed on. If consent is missing or withdrawn, the request is refused at the source — so skipping or dismissing the screen cannot result in an upload.
• Withdrawing is one tap, in Settings. It is exactly as easy as granting it, and it takes effect immediately for every future request.

3. What is collected, and how

Tavi collects only what the feature needs, and it is listed here in full:

• An anonymous account identifier. On first launch the app signs you in anonymously through Firebase Authentication. This produces a random identifier that exists only to hold your own history and subscription state together. There is no username, no email address, no password and no phone number, and the identifier is not connected to your Apple ID or to any other app.
• The screenshot you choose. Collected only when you pick an image from your photo library or take one with the camera, and only that image — Tavi cannot see the rest of your library. It is sent to our server function and on to Anthropic, and it is not written to any database or file storage at any point. It exists in the processing function's memory for the length of the request and nowhere else.
• The tone you picked and whether the image was a chat or a profile bio — sent with the image so the reply is written in the right register.
• The generated replies. The suggestions Tavi writes back are saved to your own private area of our database so the History screen can show them again, together with the date and the tone used.
• Style preferences. Which tones and phrasings you keep choosing, stored as a small profile so suggestions drift toward how you actually write. This is derived from your choices in the app, not from reading anything else on your device.
• Usage counters and subscription state. How many free analyses you have used, whether you have an active subscription and when it renews, your app language, and the date you gave or withdrew AI consent.
• Settings kept on your device only. Theme, language override and whether you have seen the introduction are stored locally and never uploaded.

Tavi does not use any advertising SDK, does not show ads, does not use an analytics or attribution SDK, and does not present an App Tracking Transparency prompt — because it does not track you across apps or websites and has nothing to ask for.

4. Every use of that data

This is the complete list of purposes. There are no others:

• To read your screenshot and write the replies — the screenshot and the tone are sent to Anthropic, which returns the suggestions.
• To show your History — the saved suggestions, dates and tones are read back to you on your own device.
• To make suggestions sound like you — the style profile biases future suggestions toward your own word choices.
• To run the free allowance and the subscription — the usage counter and entitlement decide whether an analysis is free, requires Premium, or is already covered.
• To enforce your consent decision — the consent flag and its timestamp are read before every analysis.
• To keep the service working and prevent abuse — a per-user rate limit and device attestation stop a compromised client from running up cost or flooding the AI service.
• To answer you if you write in — if you email support, we use your message and what you tell us to help.

Your data is never sold, never shared with data brokers, never used for advertising, and never used to train an AI model.

5. Who processes it, and the protection each one owes you

Tavi relies on four service providers. Each is engaged under a data-processing agreement that binds it to confidentiality, to security measures appropriate to the data, to acting only on our documented instructions, and to safeguards for international transfers — so the protection your data receives is equivalent at every provider, not only at ours.

• Anthropic PBC (United States) — the AI service that reads the screenshot and writes the replies. It receives the image and the requested tone, and nothing else. Your account identifier, email, IP-derived identity and device identifiers are deliberately stripped before the call, so the request cannot be tied back to you. Under Anthropic's commercial terms, inputs and outputs sent through its API are not used to train its models; they may be retained briefly for safety and abuse detection under Anthropic's own policy. See the Anthropic Privacy Policy.
• Google (Firebase) — Firebase Authentication for the anonymous identifier, Cloud Firestore for your history and settings, Cloud Functions for the processing described above, and Cloud Storage. Our Firestore data and our processing functions are hosted in the European Union. Google acts as our processor for these services. See the Firebase privacy and security information and the Google Privacy Policy.
• RevenueCat, Inc. (United States) — manages subscription state. It receives your anonymous account identifier and purchase information from Apple so the app knows whether Premium is active. It never receives a screenshot, a suggestion, or your name or email. See the RevenueCat Privacy Policy.
• Apple — processes the payment for any subscription through the App Store and returns only the entitlement, and provides the device attestation used to verify that requests come from a genuine copy of the app. We never see or store your payment details.

6. Conversations involving other people

A chat screenshot usually contains someone else's messages as well as yours. We take that seriously, and we handle it in three ways:

• The image is processed transiently and never retained — neither the other person's messages nor yours are stored as an image anywhere in our systems.
• Nothing sent to the AI service identifies you or them: no names from your account, no email, no phone number, no device or account identifier.
• You are responsible for what you upload. Please do not upload conversations you have no right to share, and do not use Tavi to draft messages intended to harass, deceive, threaten or impersonate anyone. See the Terms of Use.

7. What we never collect

• Your real name, email address, or phone number
• Your contacts or address book
• Your photo library beyond the single image you choose for an analysis
• Precise or approximate location
• Health data
• Payment card details
• Any advertising identifier — Tavi contains no advertising or tracking SDK at all

8. Summary of data types (Apple privacy label)

Consistent with the App Store privacy label, the data types associated with Tavi are the following. None of them are used for tracking, and none are used for advertising or marketing:

• User ID — the anonymous account identifier, linked to you, for app functionality (history, subscription, consent)
• Photos — the screenshot you select, for app functionality; processed to produce the replies and not stored
• Other user content — the generated replies and your tone preferences, linked to you, for app functionality and personalization
• Purchases — subscription status, linked to you, for app functionality
• Product interaction / diagnostics — usage counters and error logs, for app functionality and to keep the service stable

9. Legal bases (GDPR / KVKK)

For users in the EEA, the UK and Switzerland we rely on the following legal bases under the GDPR; for users in Türkiye the equivalent grounds under the KVKK (Law No. 6698) apply:

• Explicit consent (Art. 6(1)(a), and Art. 9(2)(a) where a conversation happens to reveal sensitive information) — for sending your screenshot to Anthropic. This is the consent described in section 2, and you can withdraw it at any time in Settings.
• Performance of a contract (Art. 6(1)(b)) — to provide the app, keep your history, and run your subscription.
• Legitimate interests (Art. 6(1)(f)) — to keep the service secure and available, prevent abuse, and control cost through the rate limit and device attestation.

10. Retention

• Screenshots: not retained at all. There is nothing to delete, because nothing was written.
• Your history, style profile and account record: kept until you delete them. You can remove a single analysis by swiping it away in History, or remove everything at once with Delete account in Settings.
• On-device settings: removed when you delete the app.
• Support email: kept only as long as needed to resolve your request.
• Data held by Anthropic, Google, RevenueCat and Apple is retained by them under their own published policies and any legal retention periods that apply to them.

11. Deleting your account and your data

Settings → Delete account removes your account record, your saved analyses and suggestions, your style profile and the anonymous sign-in itself, and returns the app to a clean first launch. This is immediate and cannot be undone; there is no separate request to file and no waiting period.

12. International data transfers

Our own data — your history, settings and processing functions — is hosted in the European Union. The AI processing performed by Anthropic and the subscription management performed by RevenueCat take place in the United States, and Google and Apple may process data outside your country as part of their global infrastructure. Those transfers are covered by the appropriate safeguards, including the European Commission's Standard Contractual Clauses where applicable, under the agreements described in section 5.

13. Your rights

Depending on where you live — including under the GDPR and the KVKK — you have the right to access, correct, delete, restrict or object to the processing of your personal data, to data portability, and to withdraw consent at any time without affecting processing already carried out. In practice:

• Withdraw AI consent — Settings, one tap, effective immediately.
• Delete everything — Settings → Delete account.
• Everything else — email support@oztechapps.net. Because Tavi has no account identifying you by name, please write from a device where the app is installed so we can locate the right record.

If you are in the EEA, the UK or Switzerland you also have the right to lodge a complaint with your local data protection supervisory authority; in Türkiye, with the Personal Data Protection Authority (KVKK).

14. Children

Tavi is not directed at children. You must be at least 13 years old to use it, and older where your country requires a higher age for consent to data processing. We do not knowingly collect personal information from children below that age. If you believe a child has done so, contact us and we will remove it.

15. Security

• All traffic between the app, our functions and every provider travels over encrypted (HTTPS/TLS) connections.
• The AI provider key lives only in server-side secret storage. It is never present in the app, so it cannot be extracted from a copy of the binary.
• Database rules restrict every document to its owner. Subscription status, the free-analysis counter and your saved suggestions can only be written by our server — a modified client cannot grant itself Premium or read anyone else's data.
• Requests are verified with Apple's device attestation and rate-limited per user.
• There is no password to steal, because there is no password.

16. Updates to this policy

We update this policy when there is a material change. The date at the top of this page reflects the latest revision, and the same text is shown inside the app. Continued use of the app after changes constitutes acceptance.

17. Contact

For privacy questions, requests, or to exercise any of your rights: support@oztechapps.net

Tavi is developed by OzTech App Studio in Istanbul, Türkiye.