Amma — Privacy Policy
Last updated: September 2026
Amma ("Amma Appa Arugil - AAA") is a family app that lets a parent talk
to it by voice and receive a spoken reply generated to sound like their
child, in the child's own cloned voice. To do this, Amma sends what you
say to two third-party AI services on every message — Anthropic
(whose Claude AI writes Amma's replies) and ElevenLabs
(which turns your speech into text, and text into Amma's spoken voice).
The app itself discloses this and asks you to acknowledge it during
setup, before your first conversation. This page explains what
information the app collects, how it's used, and who it's shared with in
full detail.
What we collect
- Voice recordings. When you talk to Amma, your spoken
message is sent to our server to be transcribed into text and to
generate a reply. If you set up voice cloning, a ~30 second voice
sample is sent to our voice-cloning provider (ElevenLabs) to create a
synthetic voice used for replies.
- Conversation history. Transcripts of what you say
and Amma's replies are stored on our server so the conversation can
stay contextual turn to turn.
- Family setup info. The parent's and child's names,
preferred language, and (optionally) the child's phone number, as
entered during onboarding.
- Contact names, not numbers. If you ask Amma to
call or message someone by name, that name is sent to our server so a
reply can be generated — but the actual phone number lookup happens
entirely on your device, using your phone's own Contacts. Phone
numbers themselves are never sent to or stored on our server, whether
you look one up by voice, browse them in Setup, add one from your
phonebook, or create a new one.
- Heart rate, if you connect an Apple Watch. This is
entirely optional, from the Setup tab. If connected, your current
heart rate is sent to our server along with each message you send to
Amma, so it can gently check in if it seems notably elevated or low.
We don't read any other Health data, and Amma never writes anything to
Health.
Who we share it with
- Anthropic (Claude) — processes your transcript to
generate Amma's reply text. If your question needs current information
(like today's news or a live score), Anthropic's web search also sends
a short search query — generated from your transcript, not your raw
words verbatim — to retrieve real-time results, which are then used to
write the reply.
- ElevenLabs — processes voice recordings for
speech-to-text transcription, and (if you opt in) voice cloning and
text-to-speech for replies.
- Google — the Cast SDK is used only to discover and
connect to a Chromecast/TV on your local network when you ask Amma to
play something; no personal data is shared with Google beyond standard
device discovery. The YouTube Data API is used only to search for
videos by keyword.
We do not sell your data, and we do not share it with anyone else
beyond what's listed above, for any purpose beyond what's described
here. Anthropic's own commercial terms confirm that data sent through
its API is never used to train its models. We've also opted out of
ElevenLabs' equivalent setting (as of September 2026) — audio sent from
that point on is used only to do what Amma asked of it (transcription,
or generating a reply in your child's voice), not to improve
ElevenLabs' own models.
Voice cloning consent
Voice cloning is entirely optional and requires explicit, separate
consent from whoever's voice is being recorded, given inside the app.
That consent can be withdrawn at any time from the Voice tab, and doing
so stops the cloned voice from being used for future replies.
Your choices
- You can withdraw voice-cloning consent at any time in the app.
- You can request deletion of your account's data (conversation
history, voice clone, family info) by emailing us — see below.
Data retention
Conversation history and family setup data are retained until you
request deletion. Voice clones are retained on ElevenLabs until consent
is withdrawn or deletion is requested.
Contact
Questions about this policy or a data deletion request can be sent to
dhanageetha2000@gmail.com.