Jembatan translates what you say so you can message the people you love across a
language barrier. It is built so that being honest about privacy is easy — there is simply very little to tell.
What Jembatan does with your words
When you speak, your phone converts your speech to text on the device (using your phone's built-in speech recognition).
That text is sent over an encrypted connection (HTTPS) to our translation server, which passes it to an AI language-model provider to translate and tidy it.
The translation comes back to your phone and is placed into the message field of the app you're using. You decide whether to send it.
What we do NOT do
We do not store your messages. The text you speak exists on our server only for the seconds it takes to translate it. It is never written to a database or to logs.
We do not train AI models on your messages, and we do not permit our providers to.
We never read your incoming messages. Jembatan only handles what you yourself speak into it.
We do not sell or share personal data. There is nothing to sell.
No accounts. Jembatan doesn't know your name, email, or phone number.
What we keep (metadata only)
To keep the service healthy and prevent abuse, our server keeps non-content counters:
a random installation ID (a UUID your app makes up on first run — it identifies an install, not a person, and resets if you reinstall),
how many translations that installation made on a given day, and which language pair.
These counters expire and delete themselves automatically within about 48 hours. They never contain message text.
Permissions, plainly
Microphone — to hear what you say. Speech recognition runs on-device.
Display over other apps(sideload version only) — to show the floating bubble.
Accessibility Service(sideload version only) — used for exactly one thing: placing your translated text into the message field you're typing in. It does not read your screen, collect what you browse, or monitor other apps.
Keyboard (input method) — the Jembatan Keyboard types the translation into the field. Like any keyboard it technically "handles" the field's text; it sends only your spoken text for translation and stores nothing.
Third parties
Translation requests pass through Cloudflare (our server host) and the AI model provider configured
at the time (currently OpenAI). They receive the text to translate and nothing else about you; our
agreements with them do not permit using it to train models.
Does your app collect or share any of the required user data types?
Yes (minimal)
Data types collected
Messages → Other in-app messages: spoken text, in transit only, for app functionality. Device or other IDs: a random per-install UUID, for analytics/abuse-prevention (not linked to identity).
Is data shared with third parties?
Spoken text is processed by a service provider (LLM API) solely to provide translation.
Is data processed ephemerally?
Yes — message text is processed ephemerally and never stored.
Is data encrypted in transit?
Yes (HTTPS everywhere).
Can users request data deletion?
There is no stored user data to delete; the install ID and its counters self-expire within ~48 h.
Location, contacts, financial, health, photos, files, browsing history