Privacy

What we collect Çfarë mbledhim

• The text you submit for translation and the translation we return.
• Technical metadata. Timestamp, source and target language, dialect preference, latency, request ID, the model we used.
• A salted SHA-256 hash of your IP address, used for rate limiting. We never store the raw IP.
• If you submit feedback or join the waitlist, your email and what you wrote.

How we protect you Si ju mbrojmë

• We don't sell your data. Ever.
• We don't share translation content with third parties beyond Anthropic (the LLM provider whose API we call). Per Anthropic's commercial terms, API inputs are not used to train their models, and may be retained by Anthropic for up to 30 days for abuse-monitoring purposes before deletion.
• For photo translation, image bytes are sent to Anthropic's Vision API for text extraction. We do not store the image on our side after the API call completes; the same 30-day Anthropic retention applies.
• For voice translation, the audio you record is sent to OpenAI's Whisper API for transcription. We do not store the audio bytes on our side. Per OpenAI's API data-usage policies, API inputs are not used to train their models, and may be retained by OpenAI for up to 30 days for abuse-monitoring before deletion.
• For spoken playback of translations (text-to-speech), the translated text is sent to OpenAI's TTS API to synthesize audio. The audio is cached on our server (so repeated plays of the same translation don't re-charge) and may also be retained by OpenAI for up to 30 days under the same abuse-monitoring policy. If OpenAI's TTS is unavailable, your browser's built-in speech-synthesis runs entirely on your device with nothing leaving it.
• For interpreter mode (/biseda — live two-way conversation translation), each conversation turn sends the recorded audio to OpenAI (Whisper), the resulting text to Anthropic for translation, and the translated text back to OpenAI (TTS) for spoken playback. We do not store any of the call audio on our side. The same 30-day abuse-monitoring retention disclosed above applies to OpenAI and Anthropic for these calls. Conversation transcripts appear only in your browser and are never sent to or stored on our server.
• We don't use cookies for tracking. Just a Flask session cookie for functional state.
• We don't use Google Analytics, Mixpanel or other behavioral tracking.

Retention

• Translation logs (source + translation text + metadata, including a 2-letter country code derived from Cloudflare's CF-IPCountry header — country-level only, never city or IP). Deleted after 7 days.
• Document-template translations. When you translate an official document, the extracted fields are saved so you can retrieve the result later with its reference code. These contain document details (names, dates). Deleted after 7 days.
• Cached translations (response only, keyed by content hash). Deleted after 30 days.
• Rate-limit counters. Deleted after 7 days.
• Waitlist emails. Kept until you ask us to remove them, or until we close the waitlist.

Delete my data Fshi të dhënat e mia

You can ask us to delete every record we hold that's tied to you. Submit either your email (for waitlist and feedback rows) or your IP-hash (for translation logs). If you don't have your IP-hash, send us your raw IP and we'll hash it server-side.

Limited to one deletion request per IP per day.

Contact

For data rights or privacy questions: privacy@albanisht.com.